This disclosure generally relates to devices and methods for capturing and processing images and audio from an environment of a user, and using information derived from captured images and audio.
Today, technological advancements make it possible for wearable devices to automatically capture images and audio, and store information that is associated with the captured images and audio. Certain devices have been used to digitally record aspects and personal experiences of one's life in an exercise typically called “lifelogging.” Some individuals log their life so they can retrieve moments from past activities, for example, social events, trips, etc. Lifelogging may also have significant benefits in other fields (e.g., business, fitness and healthcare, and social research). Lifelogging devices, while useful for tracking daily activities, may be improved with capability to enhance one's interaction in his environment with feedback and other advanced functionality based on the analysis of captured image and audio data.
Even though users can capture images and audio with their smartphones and some smartphone applications can process the captured information, smartphones may not be the best platform for serving as lifelogging apparatuses in view of their size and design. Lifelogging apparatuses should be small and light, so they can be easily worn. Moreover, with improvements in image capture devices, including wearable apparatuses, additional functionality may be provided to assist users in navigating in and around an environment, identifying persons and objects they encounter, and providing feedback to the users about their surroundings and activities. Therefore, there is a need for apparatuses and methods for automatically capturing and processing images and audio to provide useful information to users of the apparatuses, and for systems and methods to process and leverage information gathered by the apparatuses.
Embodiments consistent with the present disclosure provide devices and methods for automatically capturing and processing images and/or audio from an environment of a user, and systems and methods for processing information related to images and/or audio captured from the environment of the user.
In an embodiment, a system for processing audio signals is disclosed. The system may comprise: a wearable device. The wearable device may comprise a microphone having an audio sensor configured to capture the audio signals from an environment of the user and a processor. The processor may be programmed to receive the audio signals captured by the microphone; analyze the audio signals to generate a transcription; generate a summary of the transcription; cause the summary to be displayed to the user; receive a confirmation input from the user indicating that the displayed summary is correct; and cause the displayed summary to be stored.
In another embodiment, a method of processing audio signals is disclosed. The method may comprise receiving the audio signals representative of the sounds captured by a microphone from an environment of the user; analyzing the audio signals to generate a transcription; generating a summary of the transcription; causing the summary to be displayed to a user; receiving from the user a confirmation input that the displayed summary is correct; and causing the displayed summary to be stored.
Consistent with other disclosed embodiments, non-transitory computer-readable storage media may store program instructions, which are executed by at least one processor and perform any of the methods described herein.
The foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the claims.
The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate various disclosed embodiments. In the drawings:
The following detailed description refers to the accompanying drawings. Wherever possible, the same reference numbers are used in the drawings and the following description to refer to the same or similar parts. While several illustrative embodiments are described herein, modifications, adaptations and other implementations are possible. For example, substitutions, additions or modifications may be made to the components illustrated in the drawings, and the illustrative methods described herein may be modified by substituting, reordering, removing, or adding steps to the disclosed methods. Accordingly, the following detailed description is not limited to the disclosed embodiments and examples. Instead, the proper scope is defined by the appended claims. The words “comprising,” “having,” “containing,” and “including,” and other similar forms are intended to be equivalent in meaning and are open ended in that an item or items following any one of these words is not meant to be an exhaustive listing of such item or items or meant to be limited to only the listed item or items. It should also be noted that as used in the present disclosure and in the appended claims, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.
In some embodiments, apparatus 110 may communicate wirelessly or via a wire with a computing device 120. In some embodiments, computing device 120 may include, for example, a smartphone, or a tablet, or a dedicated processing unit, which may be portable (e.g., can be carried in a pocket of user 100). Although shown in
According to the disclosed embodiments, apparatus 110 may include an image sensor system 220 for capturing real-time image data of the field-of-view of user 100. In some embodiments, apparatus 110 may also include a processing unit 210 for controlling and performing the disclosed functionality of apparatus 110, such as to control the capture of image data, analyze the image data, and perform an action and/or output a feedback based on a hand-related trigger identified in the image data. According to the disclosed embodiments, a hand-related trigger may include a gesture performed by user 100 involving a portion of a hand of user 100. Further, consistent with some embodiments, a hand-related trigger may include a wrist-related trigger. Additionally, in some embodiments, apparatus 110 may include a feedback outputting unit 230 for producing an output of information to user 100.
As discussed above, apparatus 110 may include an image sensor 220 for capturing image data. The term “image sensor” refers to a device capable of detecting and converting optical signals in the near-infrared, infrared, visible, and ultraviolet spectrums into electrical signals. The electrical signals may be used to form an image or a video stream (i.e. image data) based on the detected signal. The term “image data” includes any form of data retrieved from optical signals in the near-infrared, infrared, visible, and ultraviolet spectrums. Examples of image sensors may include semiconductor charge-coupled devices (CCD), active pixel sensors in complementary metal-oxide-semiconductor (CMOS), or N-type metal-oxide-semiconductor (NMOS, Live MOS). In some cases, image sensor 220 may be part of a camera included in apparatus 110.
Apparatus 110 may also include a processor 210 for controlling image sensor 220 to capture image data and for analyzing the image data according to the disclosed embodiments. As discussed in further detail below with respect to
In some embodiments, the information or feedback information provided to user 100 may include time information. The time information may include any information related to a current time of day and, as described further below, may be presented in any sensory perceptive manner. In some embodiments, time information may include a current time of day in a preconfigured format (e.g., 2:30 pm or 14:30). Time information may include the time in the user's current time zone (e.g., based on a determined location of user 100), as well as an indication of the time zone and/or a time of day in another desired location. In some embodiments, time information may include a number of hours or minutes relative to one or more predetermined times of day. For example, in some embodiments, time information may include an indication that three hours and fifteen minutes remain until a particular hour (e.g., until 6:00 pm), or some other predetermined time. Time information may also include a duration of time passed since the beginning of a particular activity, such as the start of a meeting or the start of a jog, or any other activity. In some embodiments, the activity may be determined based on analyzed image data. In other embodiments, time information may also include additional information related to a current time and one or more other routine, periodic, or scheduled events. For example, time information may include an indication of the number of minutes remaining until the next scheduled event, as may be determined from a calendar function or other information retrieved from computing device 120 or server 250, as discussed in further detail below.
Feedback outputting unit 230 may include one or more feedback systems for providing the output of information to user 100. In the disclosed embodiments, the audible or visual feedback may be provided via any type of connected audible or visual system or both. Feedback of information according to the disclosed embodiments may include audible feedback to user 100 (e.g., using a Bluetooth™ or other wired or wirelessly connected speaker, or a bone conduction headphone). Feedback outputting unit 230 of some embodiments may additionally or alternatively produce a visible output of information to user 100, for example, as part of an augmented reality display projected onto a lens of glasses 130 or provided via a separate heads up display in communication with apparatus 110, such as a display 260 provided as part of computing device 120, which may include an onboard automobile heads up display, an augmented reality device, a virtual reality device, a smartphone, PC, table, etc.
The term “computing device” refers to a device including a processing unit and having computing capabilities. Some examples of computing device 120 include a PC, laptop, tablet, or other computing systems such as an on-board computing system of an automobile, for example, each configured to communicate directly with apparatus 110 or server 250 over network 240. Another example of computing device 120 includes a smartphone having a display 260. In some embodiments, computing device 120 may be a computing system configured particularly for apparatus 110, and may be provided integral to apparatus 110 or tethered thereto. Apparatus 110 can also connect to computing device 120 over network 240 via any known wireless standard (e.g., Wi-Fi, Bluetooth®, etc.), as well as near-filed capacitive coupling, and other short range wireless techniques, or via a wired connection. In an embodiment in which computing device 120 is a smartphone, computing device 120 may have a dedicated application installed therein. For example, user 100 may view on display 260 data (e.g., images, video clips, extracted information, feedback information, etc.) that originate from or are triggered by apparatus 110. In addition, user 100 may select part of the data for storage in server 250.
Network 240 may be a shared, public, or private network, may encompass a wide area or local area, and may be implemented through any suitable combination of wired and/or wireless communication networks. Network 240 may further comprise an intranet or the Internet. In some embodiments, network 240 may include short range or near-field wireless communication systems for enabling communication between apparatus 110 and computing device 120 provided in close proximity to each other, such as on or near a user's person, for example. Apparatus 110 may establish a connection to network 240 autonomously, for example, using a wireless module (e.g., Wi-Fi, cellular). In some embodiments, apparatus 110 may use the wireless module when being connected to an external power source, to prolong battery life. Further, communication between apparatus 110 and server 250 may be accomplished through any suitable communication channels, such as, for example, a telephone network, an extranet, an intranet, the Internet, satellite communications, off-line communications, wireless communications, transponder communications, a local area network (LAN), a wide area network (WAN), and a virtual private network (VPN).
As shown in
An example of wearable apparatus 110 incorporated with glasses 130 according to some embodiments (as discussed in connection with
In some embodiments, support 310 may include a quick release mechanism for disengaging and reengaging apparatus 110. For example, support 310 and apparatus 110 may include magnetic elements. As an alternative example, support 310 may include a male latch member and apparatus 110 may include a female receptacle. In other embodiments, support 310 can be an integral part of a pair of glasses, or sold separately and installed by an optometrist. For example, support 310 may be configured for mounting on the arms of glasses 130 near the frame front, but before the hinge. Alternatively, support 310 may be configured for mounting on the bridge of glasses 130.
In some embodiments, apparatus 110 may be provided as part of a glasses frame 130, with or without lenses. Additionally, in some embodiments, apparatus 110 may be configured to provide an augmented reality display projected onto a lens of glasses 130 (if provided), or alternatively, may include a display for projecting time information, for example, according to the disclosed embodiments. Apparatus 110 may include the additional display or alternatively, may be in communication with a separately provided display system that may or may not be attached to glasses 130.
In some embodiments, apparatus 110 may be implemented in a form other than wearable glasses, as described above with respect to
In some embodiments, apparatus 110 includes a function button 430 for enabling user 100 to provide input to apparatus 110. Function button 430 may accept different types of tactile input (e.g., a tap, a click, a double-click, a long press, a right-to-left slide, a left-to-right slide). In some embodiments, each type of input may be associated with a different action. For example, a tap may be associated with the function of taking a picture, while a right-to-left slide may be associated with the function of recording a video.
Apparatus 110 may be attached to an article of clothing (e.g., a shirt, a belt, pants, etc.), of user 100 at an edge of the clothing using a clip 431 as shown in
An example embodiment of apparatus 110 is shown in
Various views of apparatus 110 are illustrated in
The example embodiments discussed above with respect to
Processor 210, depicted in
Although, in the embodiment illustrated in
In some embodiments, processor 210 may process a plurality of images captured from the environment of user 100 to determine different parameters related to capturing subsequent images. For example, processor 210 can determine, based on information derived from captured image data, a value for at least one of the following: an image resolution, a compression ratio, a cropping parameter, frame rate, a focus point, an exposure time, an aperture size, and a light sensitivity. The determined value may be used in capturing at least one subsequent image. Additionally, processor 210 can detect images including at least one hand-related trigger in the environment of the user and perform an action and/or provide an output of information to a user via feedback outputting unit 230.
In another embodiment, processor 210 can change the aiming direction of image sensor 220. For example, when apparatus 110 is attached with clip 420, the aiming direction of image sensor 220 may not coincide with the field-of-view of user 100. Processor 210 may recognize certain situations from the analyzed image data and adjust the aiming direction of image sensor 220 to capture relevant image data. For example, in one embodiment, processor 210 may detect an interaction with another individual and sense that the individual is not fully in view, because image sensor 220 is tilted down. Responsive thereto, processor 210 may adjust the aiming direction of image sensor 220 to capture image data of the individual. Other scenarios are also contemplated where processor 210 may recognize the need to adjust an aiming direction of image sensor 220.
In some embodiments, processor 210 may communicate data to feedback-outputting unit 230, which may include any device configured to provide information to a user 100. Feedback outputting unit 230 may be provided as part of apparatus 110 (as shown) or may be provided external to apparatus 110 and communicatively coupled thereto. Feedback-outputting unit 230 may be configured to output visual or nonvisual feedback based on signals received from processor 210, such as when processor 210 recognizes a hand-related trigger in the analyzed image data.
The term “feedback” refers to any output or information provided in response to processing at least one image in an environment. In some embodiments, as similarly described above, feedback may include an audible or visible indication of time information, detected text or numerals, the value of currency, a branded product, a person's identity, the identity of a landmark or other environmental situation or condition including the street names at an intersection or the color of a traffic light, etc., as well as other information associated with each of these. For example, in some embodiments, feedback may include additional information regarding the amount of currency still needed to complete a transaction, information regarding the identified person, historical information or times and prices of admission etc. of a detected landmark etc. In some embodiments, feedback may include an audible tone, a tactile response, and/or information previously recorded by user 100. Feedback-outputting unit 230 may comprise appropriate components for outputting acoustical and tactile feedback. For example, feedback-outputting unit 230 may comprise audio headphones, a hearing aid type device, a speaker, a bone conduction headphone, interfaces that provide tactile cues, vibrotactile stimulators, etc. In some embodiments, processor 210 may communicate signals with an external feedback outputting unit 230 via a wireless transceiver 530, a wired connection, or some other communication interface. In some embodiments, feedback outputting unit 230 may also include any suitable display device for visually displaying information to user 100.
As shown in
As further shown in
Mobile power source 520 may power one or more wireless transceivers (e.g., wireless transceiver 530 in
Apparatus 110 may operate in a first processing-mode and in a second processing-mode, such that the first processing-mode may consume less power than the second processing-mode. For example, in the first processing-mode, apparatus 110 may capture images and process the captured images to make real-time decisions based on an identifying hand-related trigger, for example. In the second processing-mode, apparatus 110 may extract information from stored images in memory 550 and delete images from memory 550. In some embodiments, mobile power source 520 may provide more than fifteen hours of processing in the first processing-mode and about three hours of processing in the second processing-mode. Accordingly, different processing-modes may allow mobile power source 520 to produce sufficient power for powering apparatus 110 for various time periods (e.g., more than two hours, more than four hours, more than ten hours, etc.).
In some embodiments, apparatus 110 may use first processor 210a in the first processing-mode when powered by mobile power source 520, and second processor 210b in the second processing-mode when powered by external power source 580 that is connectable via power connector 510. In other embodiments, apparatus 110 may determine, based on predefined conditions, which processors or which processing modes to use. Apparatus 110 may operate in the second processing-mode even when apparatus 110 is not powered by external power source 580. For example, apparatus 110 may determine that it should operate in the second processing-mode when apparatus 110 is not powered by external power source 580, if the available storage space in memory 550 for storing new image data is lower than a predefined threshold.
Although one wireless transceiver is depicted in
In some embodiments, processor 210 and processor 540 are configured to extract information from captured image data. The term “extracting information” includes any process by which information associated with objects, individuals, locations, events, etc., is identified in the captured image data by any means known to those of ordinary skill in the art. In some embodiments, apparatus 110 may use the extracted information to send feedback or other real-time indications to feedback outputting unit 230 or to computing device 120. In some embodiments, processor 210 may identify in the image data the individual standing in front of user 100, and send computing device 120 the name of the individual and the last time user 100 met the individual. In another embodiment, processor 210 may identify in the image data, one or more visible triggers, including a hand-related trigger, and determine whether the trigger is associated with a person other than the user of the wearable apparatus to selectively determine whether to perform an action associated with the trigger. One such action may be to provide a feedback to user 100 via feedback-outputting unit 230 provided as part of (or in communication with) apparatus 110 or via a feedback unit 545 provided as part of computing device 120. For example, feedback-outputting unit 545 may be in communication with display 260 to cause the display 260 to visibly output information. In some embodiments, processor 210 may identify in the image data a hand-related trigger and send computing device 120 an indication of the trigger. Processor 540 may then process the received trigger information and provide an output via feedback outputting unit 545 or display 260 based on the hand-related trigger. In other embodiments, processor 540 may determine a hand-related trigger and provide suitable feedback similar to the above, based on image data received from apparatus 110. In some embodiments, processor 540 may provide instructions or other information, such as environmental information to apparatus 110 based on an identified hand-related trigger.
In some embodiments, processor 210 may identify other environmental information in the analyzed images, such as an individual standing in front user 100, and send computing device 120 information related to the analyzed information such as the name of the individual and the last time user 100 met the individual. In a different embodiment, processor 540 may extract statistical information from captured image data and forward the statistical information to server 250. For example, certain information regarding the types of items a user purchases, or the frequency a user patronizes a particular merchant, etc. may be determined by processor 540. Based on this information, server 250 may send computing device 120 coupons and discounts associated with the user's preferences.
When apparatus 110 is connected or wirelessly connected to computing device 120, apparatus 110 may transmit at least part of the image data stored in memory 550a for storage in memory 550b. In some embodiments, after computing device 120 confirms that transferring the part of image data was successful, processor 540 may delete the part of the image data. The term “delete” means that the image is marked as ‘deleted’ and other image data may be stored instead of it, but does not necessarily mean that the image data was physically removed from the memory.
As will be appreciated by a person skilled in the art having the benefit of this disclosure, numerous variations and/or modifications may be made to the disclosed embodiments. Not all components are essential for the operation of apparatus 110. Any component may be located in any appropriate apparatus and the components may be rearranged into a variety of configurations while providing the functionality of the disclosed embodiments. For example, in some embodiments, apparatus 110 may include a camera, a processor, and a wireless transceiver for sending data to another device. Therefore, the foregoing configurations are examples and, regardless of the configurations discussed above, apparatus 110 can capture, store, and/or process images.
Further, the foregoing and following description refers to storing and/or processing images or image data. In the embodiments disclosed herein, the stored and/or processed images or image data may comprise a representation of one or more images captured by image sensor 220. As the term is used herein, a “representation” of an image (or image data) may include an entire image or a portion of an image. A representation of an image (or image data) may have the same resolution or a lower resolution as the image (or image data), and/or a representation of an image (or image data) may be altered in some respect (e.g., be compressed, have a lower resolution, have one or more colors that are altered, etc.).
For example, apparatus 110 may capture an image and store a representation of the image that is compressed as a .JPG file. As another example, apparatus 110 may capture an image in color, but store a black-and-white representation of the color image. As yet another example, apparatus 110 may capture an image and store a different representation of the image (e.g., a portion of the image). For example, apparatus 110 may store a portion of an image that includes a face of a person who appears in the image, but that does not substantially include the environment surrounding the person. Similarly, apparatus 110 may, for example, store a portion of an image that includes a product that appears in the image, but does not substantially include the environment surrounding the product. As yet another example, apparatus 110 may store a representation of an image at a reduced resolution (i.e., at a resolution that is of a lower value than that of the captured image). Storing representations of images may allow apparatus 110 to save storage space in memory 550. Furthermore, processing representations of images may allow apparatus 110 to improve processing efficiency and/or help to preserve battery life.
In addition to the above, in some embodiments, any one of apparatus 110 or computing device 120, via processor 210 or 540, may further process the captured image data to provide additional functionality to recognize objects and/or gestures and/or other information in the captured image data. In some embodiments, actions may be taken based on the identified objects, gestures, or other information. In some embodiments, processor 210 or 540 may identify in the image data, one or more visible triggers, including a hand-related trigger, and determine whether the trigger is associated with a person other than the user to determine whether to perform an action associated with the trigger.
Some embodiments of the present disclosure may include an apparatus securable to an article of clothing of a user. Such an apparatus may include two portions, connectable by a connector. A capturing unit may be designed to be worn on the outside of a user's clothing, and may include an image sensor for capturing images of a user's environment. The capturing unit may be connected to or connectable to a power unit, which may be configured to house a power source and a processing device. The capturing unit may be a small device including a camera or other device for capturing images. The capturing unit may be designed to be inconspicuous and unobtrusive, and may be configured to communicate with a power unit concealed by a user's clothing. The power unit may include bulkier aspects of the system, such as transceiver antennas, at least one battery, a processing device, etc. In some embodiments, communication between the capturing unit and the power unit may be provided by a data cable included in the connector, while in other embodiments, communication may be wirelessly achieved between the capturing unit and the power unit. Some embodiments may permit alteration of the orientation of an image sensor of the capture unit, for example to better capture images of interest.
Image sensor 220 may be configured to be movable with the head of user 100 in such a manner that an aiming direction of image sensor 220 substantially coincides with a field of view of user 100. For example, as described above, a camera associated with image sensor 220 may be installed within capturing unit 710 at a predetermined angle in a position facing slightly upwards or downwards, depending on an intended location of capturing unit 710. Accordingly, the set aiming direction of image sensor 220 may match the field-of-view of user 100. In some embodiments, processor 210 may change the orientation of image sensor 220 using image data provided from image sensor 220. For example, processor 210 may recognize that a user is reading a book and determine that the aiming direction of image sensor 220 is offset from the text. That is, because the words in the beginning of each line of text are not fully in view, processor 210 may determine that image sensor 220 is tilted in the wrong direction. Responsive thereto, processor 210 may adjust the aiming direction of image sensor 220.
Orientation identification module 601 may be configured to identify an orientation of an image sensor 220 of capturing unit 710. An orientation of an image sensor 220 may be identified, for example, by analysis of images captured by image sensor 220 of capturing unit 710, by tilt or attitude sensing devices within capturing unit 710, and by measuring a relative direction of orientation adjustment unit 705 with respect to the remainder of capturing unit 710.
Orientation adjustment module 602 may be configured to adjust an orientation of image sensor 220 of capturing unit 710. As discussed above, image sensor 220 may be mounted on an orientation adjustment unit 705 configured for movement. Orientation adjustment unit 705 may be configured for rotational and/or lateral movement in response to commands from orientation adjustment module 602. In some embodiments, orientation adjustment unit 705 may be adjust an orientation of image sensor 220 via motors, electromagnets, permanent magnets, and/or any suitable combination thereof.
In some embodiments, monitoring module 603 may be provided for continuous monitoring. Such continuous monitoring may include tracking a movement of at least a portion of an object included in one or more images captured by the image sensor. For example, in one embodiment, apparatus 110 may track an object as long as the object remains substantially within the field-of-view of image sensor 220. In additional embodiments, monitoring module 603 may engage orientation adjustment module 602 to instruct orientation adjustment unit 705 to continually orient image sensor 220 towards an object of interest. For example, in one embodiment, monitoring module 603 may cause image sensor 220 to adjust an orientation to ensure that a certain designated object, for example, the face of a particular person, remains within the field-of view of image sensor 220, even as that designated object moves about. In another embodiment, monitoring module 603 may continuously monitor an area of interest included in one or more images captured by the image sensor. For example, a user may be occupied by a certain task, for example, typing on a laptop, while image sensor 220 remains oriented in a particular direction and continuously monitors a portion of each image from a series of images to detect a trigger or other event. For example, image sensor 220 may be oriented towards a piece of laboratory equipment and monitoring module 603 may be configured to monitor a status light on the laboratory equipment for a change in status, while the user's attention is otherwise occupied.
In some embodiments, consistent with the present disclosure, capturing unit 710 may include a plurality of image sensors 220. The plurality of image sensors 220 may each be configured to capture different image data. For example, when a plurality of image sensors 220 are provided, the image sensors 220 may capture images having different resolutions, may capture wider or narrower fields of view, and may have different levels of magnification. Image sensors 220 may be provided with varying lenses to permit these different configurations. In some embodiments, a plurality of image sensors 220 may include image sensors 220 having different orientations. Thus, each of the plurality of image sensors 220 may be pointed in a different direction to capture different images. The fields of view of image sensors 220 may be overlapping in some embodiments. The plurality of image sensors 220 may each be configured for orientation adjustment, for example, by being paired with an image adjustment unit 705. In some embodiments, monitoring module 603, or another module associated with memory 550, may be configured to individually adjust the orientations of the plurality of image sensors 220 as well as to turn each of the plurality of image sensors 220 on or off as may be required. In some embodiments, monitoring an object or person captured by an image sensor 220 may include tracking movement of the object across the fields of view of the plurality of image sensors 220.
Embodiments consistent with the present disclosure may include connectors configured to connect a capturing unit and a power unit of a wearable apparatus. Capturing units consistent with the present disclosure may include least one image sensor configured to capture images of an environment of a user. Power units consistent with the present disclosure may be configured to house a power source and/or at least one processing device. Connectors consistent with the present disclosure may be configured to connect the capturing unit and the power unit, and may be configured to secure the apparatus to an article of clothing such that the capturing unit is positioned over an outer surface of the article of clothing and the power unit is positioned under an inner surface of the article of clothing. Exemplary embodiments of capturing units, connectors, and power units consistent with the disclosure are discussed in further detail with respect to
Capturing unit 710 may include an image sensor 220 and an orientation adjustment unit 705 (as illustrated in
Connector 730 may include a clip 715 or other mechanical connection designed to clip or attach capturing unit 710 and power unit 720 to an article of clothing 750 as illustrated in
In some embodiments, connector 730 may include a flexible printed circuit board (PCB).
In further embodiments, an apparatus securable to an article of clothing may further include protective circuitry associated with power source 520 housed in in power unit 720.
Protective circuitry 775 may be configured to protect image sensor 220 and/or other elements of capturing unit 710 from potentially dangerous currents and/or voltages produced by mobile power source 520. Protective circuitry 775 may include passive components such as capacitors, resistors, diodes, inductors, etc., to provide protection to elements of capturing unit 710. In some embodiments, protective circuitry 775 may also include active components, such as transistors, to provide protection to elements of capturing unit 710. For example, in some embodiments, protective circuitry 775 may comprise one or more resistors serving as fuses. Each fuse may comprise a wire or strip that melts (thereby braking a connection between circuitry of image capturing unit 710 and circuitry of power unit 720) when current flowing through the fuse exceeds a predetermined limit (e.g., 500 milliamps, 900 milliamps, 1 amp, 1.1 amps, 2 amp, 2.1 amps, 3 amps, etc.) Any or all of the previously described embodiments may incorporate protective circuitry 775.
In some embodiments, the wearable apparatus may transmit data to a computing device (e.g., a smartphone, tablet, watch, computer, etc.) over one or more networks via any known wireless standard (e.g., cellular, Wi-Fi, Bluetooth®, etc.), or via near-filed capacitive coupling, other short range wireless techniques, or via a wired connection. Similarly, the wearable apparatus may receive data from the computing device over one or more networks via any known wireless standard (e.g., cellular, Wi-Fi, Bluetooth®, etc.), or via near-filed capacitive coupling, other short range wireless techniques, or via a wired connection. The data transmitted to the wearable apparatus and/or received by the wireless apparatus may include images, portions of images, identifiers related to information appearing in analyzed images or associated with analyzed audio, or any other data representing image and/or audio data. For example, an image may be analyzed and an identifier related to an activity occurring in the image may be transmitted to the computing device (e.g., the “paired device”). In the embodiments described herein, the wearable apparatus may process images and/or audio locally (on board the wearable apparatus) and/or remotely (via a computing device). Further, in the embodiments described herein, the wearable apparatus may transmit data related to the analysis of images and/or audio to a computing device for further analysis, display, and/or transmission to another device (e.g., a paired device). Further, a paired device may execute one or more applications (apps) to process, display, and/or analyze data (e.g., identifiers, text, images, audio, etc.) received from the wearable apparatus.
Some of the disclosed embodiments may involve systems, devices, methods, and software products for determining at least one keyword. For example, at least one keyword may be determined based on data collected by apparatus 110. At least one search query may be determined based on the at least one keyword. The at least one search query may be transmitted to a search engine.
In some embodiments, at least one keyword may be determined based on at least one or more images captured by image sensor 220. In some cases, the at least one keyword may be selected from a keywords pool stored in memory. In some cases, optical character recognition (OCR) may be performed on at least one image captured by image sensor 220, and the at least one keyword may be determined based on the OCR result. In some cases, at least one image captured by image sensor 220 may be analyzed to recognize: a person, an object, a location, a scene, and so forth. Further, the at least one keyword may be determined based on the recognized person, object, location, scene, etc. For example, the at least one keyword may comprise: a person's name, an object's name, a place's name, a date, a sport team's name, a movie's name, a book's name, and so forth.
In some embodiments, at least one keyword may be determined based on the user's behavior. The user's behavior may be determined based on an analysis of the one or more images captured by image sensor 220. In some embodiments, at least one keyword may be determined based on activities of a user and/or other person. The one or more images captured by image sensor 220 may be analyzed to identify the activities of the user and/or the other person who appears in one or more images captured by image sensor 220. In some embodiments, at least one keyword may be determined based on at least one or more audio segments captured by apparatus 110. In some embodiments, at least one keyword may be determined based on at least GPS information associated with the user. In some embodiments, at least one keyword may be determined based on at least the current time and/or date.
In some embodiments, at least one search query may be determined based on at least one keyword. In some cases, the at least one search query may comprise the at least one keyword. In some cases, the at least one search query may comprise the at least one keyword and additional keywords provided by the user. In some cases, the at least one search query may comprise the at least one keyword and one or more images, such as images captured by image sensor 220. In some cases, the at least one search query may comprise the at least one keyword and one or more audio segments, such as audio segments captured by apparatus 110.
In some embodiments, the at least one search query may be transmitted to a search engine. In some embodiments, search results provided by the search engine in response to the at least one search query may be provided to the user. In some embodiments, the at least one search query may be used to access a database.
For example, in one embodiment, the keywords may include a name of a type of food, such as quinoa, or a brand name of a food product; and the search will output information related to desirable quantities of consumption, facts about the nutritional profile, and so forth. In another example, in one embodiment, the keywords may include a name of a restaurant, and the search will output information related to the restaurant, such as a menu, opening hours, reviews, and so forth. The name of the restaurant may be obtained using OCR on an image of signage, using GPS information, and so forth. In another example, in one embodiment, the keywords may include a name of a person, and the search will provide information from a social network profile of the person. The name of the person may be obtained using OCR on an image of a name tag attached to the person's shirt, using face recognition algorithms, and so forth. In another example, in one embodiment, the keywords may include a name of a book, and the search will output information related to the book, such as reviews, sales statistics, information regarding the author of the book, and so forth. In another example, in one embodiment, the keywords may include a name of a movie, and the search will output information related to the movie, such as reviews, box office statistics, information regarding the cast of the movie, show times, and so forth. In another example, in one embodiment, the keywords may include a name of a sport team, and the search will output information related to the sport team, such as statistics, latest results, future schedule, information regarding the players of the sport team, and so forth. For example, the name of the sport team may be obtained using audio recognition algorithms.
Camera-Based Directional Hearing Aid
As discussed previously, the disclosed embodiments may include providing feedback, such as acoustical and tactile feedback, to one or more auxiliary devices in response to processing at least one image in an environment. In some embodiments, the auxiliary device may be an earpiece or other device used to provide auditory feedback to the user, such as a hearing aid. Traditional hearing aids often use microphones to amplify sounds in the user's environment. These traditional systems, however, are often unable to distinguish between sounds that may be of particular importance to the wearer of the device, or may do so on a limited basis. Using the systems and methods of the disclosed embodiments, various improvements to traditional hearing aids are provided, as described in detail below.
In one embodiment, a camera-based directional hearing aid may be provided for selectively amplifying sounds based on a look direction of a user. The hearing aid may communicate with an image capturing device, such as apparatus 110, to determine the look direction of the user. This look direction may be used to isolate and/or selectively amplify sounds received from that direction (e.g., sounds from individuals in the user's look direction, etc.). Sounds received from directions other than the user's look direction may be suppressed, attenuated, filtered or the like.
Hearing interface device 1710 may be any device configured to provide audible feedback to user 100. Hearing interface device 1710 may correspond to feedback outputting unit 230, described above, and therefore any descriptions of feedback outputting unit 230 may also apply to hearing interface device 1710. In some embodiments, hearing interface device 1710 may be separate from feedback outputting unit 230 and may be configured to receive signals from feedback outputting unit 230. As shown in
Hearing interface device 1710 may have various other configurations or placement locations. In some embodiments, hearing interface device 1710 may comprise a bone conduction headphone 1711, as shown in
Apparatus 110 may be configured to determine a user look direction 1750 of user 100. In some embodiments, user look direction 1750 may be tracked by monitoring a direction of the chin, or another body part or face part of user 100 relative to an optical axis of a camera sensor 1751. Apparatus 110 may be configured to capture one or more images of the surrounding environment of user, for example, using image sensor 220. The captured images may include a representation of a chin of user 100, which may be used to determine user look direction 1750. Processor 210 (and/or processors 210a and 210b) may be configured to analyze the captured images and detect the chin or another part of user 100 using various image detection or processing algorithms (e.g., using convolutional neural networks (CNN), scale-invariant feature transform (SIFT), histogram of oriented gradients (HOG) features, or other techniques). Based on the detected representation of a chin of user 100, look direction 1750 may be determined. Look direction 1750 may be determined in part by comparing the detected representation of a chin of user 100 to an optical axis of a camera sensor 1751. For example, the optical axis 1751 may be known or fixed in each image and processor 210 may determine look direction 1750 by comparing a representative angle of the chin of user 100 to the direction of optical axis 1751. While the process is described using a representation of a chin of user 100, various other features may be detected for determining user look direction 1750, including the user's face, nose, eyes, hand, etc.
In other embodiments, user look direction 1750 may be aligned more closely with the optical axis 1751. For example, as discussed above, apparatus 110 may be affixed to a pair of glasses of user 100, as shown in
Apparatus 110 may further comprise one or more microphones 1720 for capturing sounds from the environment of user 100. Microphone 1720 may also be configured to determine a directionality of sounds in the environment of user 100. For example, microphone 1720 may comprise one or more directional microphones, which may be more sensitive to picking up sounds in certain directions. For example, microphone 1720 may comprise a unidirectional microphone, designed to pick up sound from a single direction or small range of directions. Microphone 1720 may also comprise a cardioid microphone, which may be sensitive to sounds from the front and sides. Microphone 1720 may also include a microphone array, which may comprise additional microphones, such as microphone 1721 on the front of apparatus 110, or microphone 1722, placed on the side of apparatus 110. In some embodiments, microphone 1720 may be a multi-port microphone for capturing multiple audio signals. The microphones shown in
As a preliminary step before other audio analysis operations, the sound captured from an environment of a user may be classified using any audio classification technique. For example, the sound may be classified into segments containing music, tones, laughter, screams, or the like. Indications of the respective segments may be logged in a database and may prove highly useful for life logging applications. As one example, the logged information may enable the system to to retrieve and/or determine a mood when the user met another person. Additionally, such processing is relatively fast and efficient, and does not require significant computing resources, and transmitting the information to a destination does not require significant bandwidth. Moreover, once certain parts of the audio are classified as non-speech, more computing resources may be available for processing the other segments.
Based on the determined user look direction 1750, processor 210 may selectively condition or amplify sounds from a region associated with user look direction 1750.
Processor 210 may be configured to cause selective conditioning of sounds in the environment of user 100 based on region 1830. The conditioned audio signal may be transmitted to hearing interface device 1710, and thus may provide user 100 with audible feedback corresponding to the look direction of the user. For example, processor 210 may determine that sound 1820 (which may correspond to the voice of an individual 1810, or to noise for example) is within region 1830. Processor 210 may then perform various conditioning techniques on the audio signals received from microphone 1720. The conditioning may include amplifying audio signals determined to correspond to sound 1820 relative to other audio signals. Amplification may be accomplished digitally, for example by processing audio signals associated with 1820 relative to other signals. Amplification may also be accomplished by changing one or more parameters of microphone 1720 to focus on audio sounds emanating from region 1830 (e.g., a region of interest) associated with user look direction 1750. For example, microphone 1720 may be a directional microphone that and processor 210 may perform an operation to focus microphone 1720 on sound 1820 or other sounds within region 1830. Various other techniques for amplifying sound 1820 may be used, such as using a beamforming microphone array, acoustic telescope techniques, etc.
Conditioning may also include attenuation or suppressing one or more audio signals received from directions outside of region 1830. For example, processor 210 may attenuate sounds 1821 and 1822. Similar to amplification of sound 1820, attenuation of sounds may occur through processing audio signals, or by varying one or more parameters associated with one or more microphones 1720 to direct focus away from sounds emanating from outside of region 1830.
In some embodiments, conditioning may further include changing a tone of audio signals corresponding to sound 1820 to make sound 1820 more perceptible to user 100. For example, user 100 may have lesser sensitivity to tones in a certain range and conditioning of the audio signals may adjust the pitch of sound 1820 to make it more perceptible to user 100. For example, user 100 may experience hearing loss in frequencies above 10 khz. Accordingly, processor 210 may remap higher frequencies (e.g., at 15 khz) to 10 khz. In some embodiments, processor 210 may be configured to change a rate of speech associated with one or more audio signals. Accordingly, processor 210 may be configured to detect speech within one or more audio signals received by microphone 1720, for example using voice activity detection (VAD) algorithms or techniques. If sound 1820 is determined to correspond to voice or speech, for example from individual 1810, processor 210 may be configured to vary the playback rate of sound 1820. For example, the rate of speech of individual 1810 may be decreased to make the detected speech more perceptible to user 100. Various other processing may be performed, such as modifying the tone of sound 1820 to maintain the same pitch as the original audio signal, or to reduce noise within the audio signal. If speech recognition has been performed on the audio signal associated with sound 1820, conditioning may further include modifying the audio signal based on the detected speech. For example, processor 210 may introduce pauses or increase the duration of pauses between words and/or sentences, which may make the speech easier to understand.
The conditioned audio signal may then be transmitted to hearing interface device 1710 and produced for user 100. Thus, in the conditioned audio signal, sound 1820 may be easier to hear to user 100, louder and/or more easily distinguishable than sounds 1821 and 1822, which may represent background noise within the environment.
In step 1910, process 1900 may include receiving a plurality of images from an environment of a user captured by a camera. The camera may be a wearable camera such as camera 1730 of apparatus 110. In step 1912, process 1900 may include receiving audio signals representative of sounds received by at least one microphone. The microphone may be configured to capture sounds from an environment of the user. For example, the microphone may be microphone 1720, as described above. Accordingly, the microphone may include a directional microphone, a microphone array, a multi-port microphone, or various other types of microphones. In some embodiments, the microphone and wearable camera may be included in a common housing, such as the housing of apparatus 110. The one or more processors performing process 1900 may also be included in the housing or may be included in a second housing. In such embodiments, the processor(s) may be configured to receive images and/or audio signals from the common housing via a wireless link (e.g., Bluetooth™, NFC, etc.). Accordingly, the common housing (e.g., apparatus 110) and the second housing (e.g., computing device 120) may further comprise transmitters or various other communication components.
In step 1914, process 1900 may include determining a look direction for the user based on analysis of at least one of the plurality of images. As discussed above, various techniques may be used to determine the user look direction. In some embodiments, the look direction may be determined based, at least in part, upon detection of a representation of a chin of a user in one or more images. The images may be processed to determine a pointing direction of the chin relative to an optical axis of the wearable camera, as discussed above.
In step 1916, process 1900 may include causing selective conditioning of at least one audio signal received by the at least one microphone from a region associated with the look direction of the user. As described above, the region may be determined based on the user look direction determined in step 1914. The range may be associated with an angular width about the look direction (e.g., 10 degrees, 20 degrees, 45 degrees, etc.). Various forms of conditioning may be performed on the audio signal, as discussed above. In some embodiments, conditioning may include changing the tone or playback speed of an audio signal. For example, conditioning may include changing a rate of speech associated with the audio signal. In some embodiments, the conditioning may include amplification of the audio signal relative to other audio signals received from outside of the region associated with the look direction of the user. Amplification may be performed by various means, such as operation of a directional microphone configured to focus on audio sounds emanating from the region, or varying one or more parameters associated with the microphone to cause the microphone to focus on audio sounds emanating from the region. The amplification may include attenuating or suppressing one or more audio signals received by the microphone from directions outside the region associated with the look direction of user 100.
In step 1918, process 1900 may include causing transmission of the at least one conditioned audio signal to a hearing interface device configured to provide sound to an ear of the user. The conditioned audio signal, for example, may be transmitted to hearing interface device 1710, which may provide sound corresponding to the audio signal to user 100. The processor performing process 1900 may further be configured to cause transmission to the hearing interface device of one or more audio signals representative of background noise, which may be attenuated relative to the at least one conditioned audio signal. For example, processor 210 may be configured to transmit audio signals corresponding to sounds 1820, 1821, and 1822. The signal associated with 1820, however, may be modified in a different manner, for example amplified, from sounds 1821 and 1822 based on a determination that sound 1820 is within region 1830. In some embodiments, hearing interface device 1710 may include a speaker associated with an earpiece. For example, hearing interface device may be inserted at least partially into the ear of the user for providing audio to the user. Hearing interface device may also be external to the ear, such as a behind-the-ear hearing device, one or more headphones, a small portable speaker, or the like. In some embodiments, hearing interface device may include a bone conduction microphone, configured to provide an audio signal to user through vibrations of a bone of the user's head. Such devices may be placed in contact with the exterior of the user's skin, or may be implanted surgically and attached to the bone of the user.
Hearing Aid with Voice and/or Image Recognition
Consistent with the disclosed embodiments, a hearing aid may selectively amplify audio signals associated with a voice of a recognized individual. The hearing aid system may store voice characteristics and/or facial features of a recognized person to aid in recognition and selective amplification. For example, when an individual enters the field of view of apparatus 110, the individual may be recognized as an individual that has been introduced to the device, or that has possibly interacted with user 100 in the past (e.g., a friend, colleague, relative, prior acquaintance, etc.). Accordingly, audio signals associated with the recognized individual's voice may be isolated and/or selectively amplified relative to other sounds in the environment of the user. Audio signals associated with sounds received from directions other than the individual's direction may be suppressed, attenuated, filtered or the like.
User 100 may wear a hearing aid device similar to the camera-based hearing aid device discussed above. For example, the hearing aid device may be hearing interface device 1720, as shown in
In some embodiments, hearing interface device 1710 may comprise a bone conduction headphone 1711, as shown in
Hearing interface device 1710 may be configured to communicate with a camera device, such as apparatus 110. Such communication may be through a wired connection, or may be made wirelessly (e.g., using a Bluetooth™, NFC, or forms of wireless communication). As discussed above, apparatus 110 may be worn by user 100 in various configurations, including being physically connected to a shirt, necklace, a belt, glasses, a wrist strap, a button, or other articles associated with user 100. In some embodiments, one or more additional devices may also be included, such as computing device 120. Accordingly, one or more of the processes or functions described herein with respect to apparatus 110 or processor 210 may be performed by computing device 120 and/or processor 540.
As discussed above, apparatus 110 may comprise at least one microphone and at least one image capture device. Apparatus 110 may comprise microphone 1720, as described with respect to
Apparatus 110 may be configured to recognize an individual in the environment of user 100.
Facial recognition component 2040 may be configured to identify one or more faces within the environment of user 100. For example, facial recognition component 2040 may identify facial features on the face 2011 of individual 2010, such as the eyes, nose, cheekbones, jaw, or other features. Facial recognition component 2040 may then analyze the relative size and position of these features to identify the user. Facial recognition component 2040 may utilize one or more algorithms for analyzing the detected features, such as principal component analysis (e.g., using eigenfaces), linear discriminant analysis, elastic bunch graph matching (e.g., using Fisherface), Local Binary Patterns Histograms (LBPH), Scale Invariant Feature Transform (SIFT), Speed Up Robust Features (SURF), or the like. Other facial recognition techniques such as 3-Dimensional recognition, skin texture analysis, and/or thermal imaging may also be used to identify individuals. Other features besides facial features may also be used for identification, such as the height, body shape, or other distinguishing features of individual 2010.
Facial recognition component 2040 may access a database or data associated with user 100 to determine if the detected facial features correspond to a recognized individual. For example, a processor 210 may access a database 2050 containing information about individuals known to user 100 and data representing associated facial features or other identifying features. Such data may include one or more images of the individuals, or data representative of a face of the user that may be used for identification through facial recognition. Database 2050 may be any device capable of storing information about one or more individuals, and may include a hard drive, a solid state drive, a web storage platform, a remote server, or the like. Database 2050 may be located within apparatus 110 (e.g., within memory 550) or external to apparatus 110, as shown in
In some embodiments, user 100 may have access to database 2050, such as through a web interface, an application on a mobile device, or through apparatus 110 or an associated device. For example, user 100 may be able to select which contacts are recognizable by apparatus 110 and/or delete or add certain contacts manually. In some embodiments, a user or administrator may be able to train facial recognition component 2040. For example, user 100 may have an option to confirm or reject identifications made by facial recognition component 2040, which may improve the accuracy of the system. This training may occur in real time, as individual 2010 is being recognized, or at some later time.
Other data or information may also inform the facial identification process. In some embodiments, processor 210 may use various techniques to recognize the voice of individual 2010, as described in further detail below. The recognized voice pattern and the detected facial features may be used, either alone or in combination, to determine that individual 2010 is recognized by apparatus 110. Processor 210 may also determine a user look direction 1750, as described above, which may be used to verify the identity of individual 2010. For example, if user 100 is looking in the direction of individual 2010 (especially for a prolonged period), this may indicate that individual 2010 is recognized by user 100, which may be used to increase the confidence of facial recognition component 2040 or other identification means.
Processor 210 may further be configured to determine whether individual 2010 is recognized by user 100 based on one or more detected audio characteristics of sounds associated with a voice of individual 2010. Returning to
In some embodiments, apparatus 110 may detect the voice of an individual that is not within the field of view of apparatus 110. For example, the voice may be heard over a speakerphone, from a back seat, or the like. In such embodiments, recognition of an individual may be based on the voice of the individual only, in the absence of a speaker in the field of view. Processor 210 may analyze the voice of the individual as described above, for example, by determining whether the detected voice matches a voiceprint of an individual in database 2050.
After determining that individual 2010 is a recognized individual of user 100, processor 210 may cause selective conditioning of audio associated with the recognized individual. The conditioned audio signal may be transmitted to hearing interface device 1710, and thus may provide user 100 with audio conditioned based on the recognized individual. For example, the conditioning may include amplifying audio signals determined to correspond to sound 2020 (which may correspond to voice 2012 of individual 2010) relative to other audio signals. In some embodiments, amplification may be accomplished digitally, for example by processing audio signals associated with sound 2020 relative to other signals. Additionally, or alternatively, amplification may be accomplished by changing one or more parameters of microphone 1720 to focus on audio sounds associated with individual 2010. For example, microphone 1720 may be a directional microphone and processor 210 may perform an operation to focus microphone 1720 on sound 2020. Various other techniques for amplifying sound 2020 may be used, such as using a beamforming microphone array, acoustic telescope techniques, etc.
In some embodiments, selective conditioning may include attenuation or suppressing one or more audio signals received from directions not associated with individual 2010. For example, processor 210 may attenuate sounds 2021 and/or 2022. Similar to amplification of sound 2020, attenuation of sounds may occur through processing audio signals, or by varying one or more parameters associated with microphone 1720 to direct focus away from sounds not associated with individual 2010.
Selective conditioning may further include determining whether individual 2010 is speaking. For example, processor 210 may be configured to analyze images or videos containing representations of individual 2010 to determine when individual 2010 is speaking, for example, based on detected movement of the recognized individual's lips. This may also be determined through analysis of audio signals received by microphone 1720, for example by detecting the voice 2012 of individual 2010. In some embodiments, the selective conditioning may occur dynamically (initiated and/or terminated) based on whether or not the recognized individual is speaking.
In some embodiments, conditioning may further include changing a tone of one or more audio signals corresponding to sound 2020 to make the sound more perceptible to user 100. For example, user 100 may have lesser sensitivity to tones in a certain range and conditioning of the audio signals may adjust the pitch of sound 2020. In some embodiments, processor 210 may be configured to change a rate of speech associated with one or more audio signals. For example, sound 2020 may be determined to correspond to voice 2012 of individual 2010. Processor 210 may be configured to vary the rate of speech of individual 2010 to make the detected speech more perceptible to user 100. Various other processing may be performed, such as modifying the tone of sound 2020 to maintain the same pitch as the original audio signal, or to reduce noise within the audio signal.
In some embodiments, processor 210 may determine a region 2030 associated with individual 2010. Region 2030 may be associated with a direction of individual 2010 relative to apparatus 110 or user 100. The direction of individual 2010 may be determined using camera 1730 and/or microphone 1720 using the methods described above. As shown in
The conditioned audio signal may then be transmitted to hearing interface device 1710 and produced for user 100. Thus, in the conditioned audio signal, sound 2020 (and specifically voice 2012) may be louder and/or more easily distinguishable than sounds 2021 and 2022, which may represent background noise within the environment.
In some embodiments, processor 210 may perform further analysis based on captured images or videos to determine how to selectively condition audio signals associated with a recognized individual. In some embodiments, processor 210 may analyze the captured images to selectively condition audio associated with one individual relative to others. For example, processor 210 may determine the direction of a recognized individual relative to the user based on the images and may determine how to selectively condition audio signals associated with the individual based on the direction. If the recognized individual is standing to the front of the user, audio associated with that user may be amplified (or otherwise selectively conditioned) relative to audio associated with an individual standing to the side of the user. Similarly, processor 210 may selectively condition audio signals associated with an individual based on proximity to the user. Processor 210 may determine a distance from the user to each individual based on captured images and may selectively condition audio signals associated with the individuals based on the distance. For example, an individual closer to the user may be prioritized higher than an individual that is farther away.
In some embodiments, selective conditioning of audio signals associated with a recognized individual may be based on the identities of individuals within the environment of the user. For example, where multiple individuals are detected in the images, processor 210 may use one or more facial recognition techniques to identify the individuals, as described above. Audio signals associated with individuals that are known to user 100 may be selectively amplified or otherwise conditioned to have priority over unknown individuals. For example, processor 210 may be configured to attenuate or silence audio signals associated with bystanders in the user's environment, such as a noisy office mate, etc. In some embodiments, processor 210 may also determine a hierarchy of individuals and give priority based on the relative status of the individuals. This hierarchy may be based on the individual's position within a family or an organization (e.g., a company, sports team, club, etc.) relative to the user. For example, the user's boss may be ranked higher than a co-worker or a member of the maintenance staff and thus may have priority in the selective conditioning process. In some embodiments, the hierarchy may be determined based on a list or database. Individuals recognized by the system may be ranked individually or grouped into tiers of priority. This database may be maintained specifically for this purpose, or may be accessed externally. For example, the database may be associated with a social network of the user (e.g., Facebook™, LinkedIn™, etc.) and individuals may be prioritized based on their grouping or relationship with the user. Individuals identified as “close friends” or family, for example, may be prioritized over acquaintances of the user.
Selective conditioning may be based on a determined behavior of one or more individuals determined based on the captured images. In some embodiments, processor 210 may be configured to determine a look direction of the individuals in the images. Accordingly, the selective conditioning may be based on behavior of the other individuals towards the recognized individual. For example, processor 210 may selectively condition audio associated with a first individual that one or more other users are looking at. If the attention of the individuals shifts to a second individual, processor 210 may then switch to selectively condition audio associated with the second user. In some embodiments, processor 210 may be configured to selectively condition audio based on whether a recognized individual is speaking to the user or to another individual. For example, when the recognized individual is speaking to the user, the selective conditioning may include amplifying an audio signal associated with the recognized individual relative to other audio signals received from directions outside a region associated with the recognized individual. When the recognized individual is speaking to another individual, the selective conditioning may include attenuating the audio signal relative to other audio signals received from directions outside the region associated with the recognized individual.
In some embodiments, processor 210 may have access to one or more voiceprints of individuals, which may facilitate selective conditioning of voice 2012 of individual 2010 in relation to other sounds or voices. Having a speaker's voiceprint, and a high quality voiceprint in particular, may provide for fast and efficient speaker separation. A high quality voice print may be collected, for example, when the user speaks alone, preferably in a quiet environment. By having a voiceprint of one or more speakers, it is possible to separate an ongoing voice signal almost in real time, e.g. with a minimal delay, using a sliding time window. The delay may be, for example 10 ms, 20 ms, 30 ms, 50 ms, 100 ms, or the like. Different time windows may be selected, depending on the quality of the voice print, on the quality of the captured audio, the difference in characteristics between the speaker and other speaker(s), the available processing resources, the required separation quality, or the like. In some embodiments, a voice print may be extracted from a segment of a conversation in which an individual speaks alone, and then used for separating the individual's voice later in the conversation, whether the individual's is recognized or not.
Separating voices may be performed as follows: spectral features, also referred to as spectral attributes, spectral envelope, or spectrogram may be extracted from a clean audio of a single speaker and fed into a pre-trained first neural network, which generates or updates a signature of the speaker's voice based on the extracted features. The audio may be for example, of one second of clean voice. The output signature may be a vector representing the speaker's voice, such that the distance between the vector and another vector extracted from the voice of the same speaker is typically smaller than the distance between the vector and a vector extracted from the voice of another speaker. The speaker's model may be pre-generated from a captured audio. Alternatively or additionally, the model may be generated after a segment of the audio in which only the speaker speaks, followed by another segment in which the speaker and another speaker (or background noise) is heard, and which it is required to separate.
Then, to separate the speaker's voice from additional speakers or background noise in a noisy audio, a second pre-trained neural network may receive the noisy audio and the speaker's signature and output an audio (which may also be represented as attributes) of the voice of the speaker as extracted from the noisy audio, separated from the other speech or background noise. It will be appreciated that the same or additional neural networks may be used to separate the voices of multiple speakers. For example, if there are two possible speakers, two neural networks may be activated, each with models of the same noisy output and one of the two speakers. Alternatively, a neural network may receive voice signatures of two or more speakers and output the voice of each of the speakers separately. Accordingly, the system may generate two or more different audio outputs, each comprising the speech of the respective speaker. In some embodiments, if separation is impossible, the input voice may only be cleaned from background noise.
In step 2110, process 2100 may include receiving a plurality of images from an environment of a user captured by a camera. The images may be captured by a wearable camera such as camera 1730 of apparatus 110. In step 2112, process 2100 may include identifying a representation of a recognized individual in at least one of the plurality of images. Individual 2010 may be recognized by processor 210 using facial recognition component 2040, as described above. For example, individual 2010 may be a friend, colleague, relative, or prior acquaintance of the user. Processor 210 may determine whether an individual represented in at least one of the plurality of images is a recognized individual based on one or more detected facial features associated with the individual. Processor 210 may also determine whether the individual is recognized based on one or more detected audio characteristics of sounds determined to be associated with a voice of the individual, as described above.
In step 2114, process 2100 may include receiving audio signals representative of sounds captured by a microphone. For example, apparatus 110 may receive audio signals representative of sounds 2020, 2021, and 2022, captured by microphone 1720. Accordingly, the microphone may include a directional microphone, a microphone array, a multi-port microphone, or various other types of microphones, as described above. In some embodiments, the microphone and wearable camera may be included in a common housing, such as the housing of apparatus 110. The one or more processors performing process 2100 may also be included in the housing (e.g., processor 210), or may be included in a second housing. Where a second housing is used, the processor(s) may be configured to receive images and/or audio signals from the common housing via a wireless link (e.g., Bluetooth™, NFC, etc.). Accordingly, the common housing (e.g., apparatus 110) and the second housing (e.g., computing device 120) may further comprise transmitters, receivers, and/or various other communication components.
In step 2116, process 2100 may include cause selective conditioning of at least one audio signal received by the at least one microphone from a region associated with the at least one recognized individual. As described above, the region may be determined based on a determined direction of the recognized individual based one or more of the plurality of images or audio signals. The range may be associated with an angular width about the direction of the recognized individual (e.g., 10 degrees, 20 degrees, 45 degrees, etc.).
Various forms of conditioning may be performed on the audio signal, as discussed above. In some embodiments, conditioning may include changing the tone or playback speed of an audio signal. For example, conditioning may include changing a rate of speech associated with the audio signal. In some embodiments, the conditioning may include amplification of the audio signal relative to other audio signals received from outside of the region associated with the recognized individual. Amplification may be performed by various means, such as operation of a directional microphone configured to focus on audio sounds emanating from the region or varying one or more parameters associated with the microphone to cause the microphone to focus on audio sounds emanating from the region. The amplification may include attenuating or suppressing one or more audio signals received by the microphone from directions outside the region. In some embodiments, step 2116 may further comprise determining, based on analysis of the plurality of images, that the recognized individual is speaking and trigger the selective conditioning based on the determination that the recognized individual is speaking. For example, the determination that the recognized individual is speaking may be based on detected movement of the recognized individual's lips. In some embodiments, selective conditioning may be based on further analysis of the captured images as described above, for example, based on the direction or proximity of the recognized individual, the identity of the recognized individual, the behavior of other individuals, etc.
In step 2118, process 2100 may include causing transmission of the at least one conditioned audio signal to a hearing interface device configured to provide sound to an ear of the user. The conditioned audio signal, for example, may be transmitted to hearing interface device 1710, which may provide sound corresponding to the audio signal to user 100. The processor performing process 2100 may further be configured to cause transmission to the hearing interface device of one or more audio signals representative of background noise, which may be attenuated relative to the at least one conditioned audio signal. For example, processor 210 may be configured to transmit audio signals corresponding to sounds 2020, 2021, and 2022. The signal associated with 2020, however, may be amplified in relation to sounds 2021 and 2022 based on a determination that sound 2020 is within region 2030. In some embodiments, hearing interface device 1710 may include a speaker associated with an earpiece. For example, hearing interface device 1710 may be inserted at least partially into the ear of the user for providing audio to the user. Hearing interface device may also be external to the ear, such as a behind-the-ear hearing device, one or more headphones, a small portable speaker, or the like. In some embodiments, hearing interface device may include a bone conduction microphone, configured to provide an audio signal to user through vibrations of a bone of the user's head. Such devices may be placed in contact with the exterior of the user's skin, or may be implanted surgically and attached to the bone of the user.
In addition to recognizing voices of individuals speaking to user 100, the systems and methods described above may also be used to recognize the voice of user 100. For example, voice recognition unit 2041 may be configured to analyze audio signals representative of sounds collected from the user's environment to recognize the voice of user 100. Similar to the selective conditioning of the voice of recognized individuals, the voice of user 100 may be selectively conditioned. For example, sounds may be collected by microphone 1720, or by a microphone of another device, such as a mobile phone (or a device linked to a mobile phone). Audio signals corresponding to the voice of user 100 may be selectively transmitted to a remote device, for example, by amplifying the voice of user 100 and/or attenuating or eliminating altogether sounds other than the user's voice. Accordingly, a voiceprint of one or more users of apparatus 110 may be collected and/or stored to facilitate detection and/or isolation of the user's voice, as described in further detail above.
In step 2210, process 2200 may include receiving audio signals representative of sounds captured by a microphone. For example, apparatus 110 may receive audio signals representative of sounds 2020, 2021, and 2022, captured by microphone 1720. Accordingly, the microphone may include a directional microphone, a microphone array, a multi-port microphone, or various other types of microphones, as described above. In step 2212, process 2200 may include identifying, based on analysis of the received audio signals, one or more voice audio signals representative of a recognized voice of the user. For example, the voice of the user may be recognized based on a voiceprint associated with the user, which may be stored in memory 550, database 2050, or other suitable locations. Processor 210 may recognize the voice of the user, for example, using voice recognition component 2041. Processor 210 may separate an ongoing voice signal associated with the user almost in real time, e.g. with a minimal delay, using a sliding time window. The voice may be separated by extracting spectral features of an audio signal according to the methods described above.
In step 2214, process 2200 may include causing transmission, to a remotely located device, of the one or more voice audio signals representative of the recognized voice of the user. The remotely located device may be any device configured to receive audio signals remotely, either by a wired or wireless form of communication. In some embodiments, the remotely located device may be another device of the user, such as a mobile phone, an audio interface device, or another form of computing device. In some embodiments, the voice audio signals may be processed by the remotely located device and/or transmitted further. In step 2216, process 2200 may include preventing transmission, to the remotely located device, of at least one background noise audio signal different from the one or more voice audio signals representative of a recognized voice of the user. For example, processor 210 may attenuate and/or eliminate audio signals associated with sounds 2020, 2021, or 2023, which may represent background noise. The voice of the user may be separated from other noises using the audio processing techniques described above.
In an exemplary illustration, the voice audio signals may be captured by a headset or other device worn by the user. The voice of the user may be recognized and isolated from the background noise in the environment of the user. The headset may transmit the conditioned audio signal of the user's voice to a mobile phone of the user. For example, the user may be on a telephone call and the conditioned audio signal may be transmitted by the mobile phone to a recipient of the call. The voice of the user may also be recorded by the remotely located device. The audio signal, for example, may be stored on a remote server or other computing device. In some embodiments, the remotely located device may process the received audio signal, for example, to convert the recognized user's voice into text.
Lip-Tracking Hearing Aid
Consistent with the disclosed embodiments, a hearing aid system may selectively amplify audio signals based on tracked lip movements. The hearing aid system analyzes captured images of the environment of a user to detect lips of an individual and track movement of the individual's lips. The tracked lip movements may serve as a cue for selectively amplifying audio received by the hearing aid system. For example, voice signals determined to sync with the tracked lip movements or that are consistent with the tracked lip movements may be selectively amplified or otherwise conditioned. Audio signals that are not associated with the detected lip movement may be suppressed, attenuated, filtered or the like.
User 100 may wear a hearing aid device consistent with the camera-based hearing aid device discussed above. For example, the hearing aid device may be hearing interface device 1710, as shown in
In some embodiments, hearing interface device 1710 may comprise a bone conduction headphone 1711, as shown in
Hearing interface device 1710 may be configured to communicate with a camera device, such as apparatus 110. Such communication may be through a wired connection, or may be made wirelessly (e.g., using a Bluetooth™, NFC, or forms of wireless communication). As discussed above, apparatus 110 may be worn by user 100 in various configurations, including being physically connected to a shirt, necklace, a belt, glasses, a wrist strap, a button, or other articles associated with user 100. In some embodiments, one or more additional devices may also be included, such as computing device 120. Accordingly, one or more of the processes or functions described herein with respect to apparatus 110 or processor 210 may be performed by computing device 120 and/or processor 540.
As discussed above, apparatus 110 may comprise at least one microphone and at least one image capture device. Apparatus 110 may comprise microphone 1720, as described with respect to
Processor 210 (and/or processors 210a and 210b) may be configured to detect a mouth and/or lips associated with an individual within the environment of user 100.
In some embodiments, processor 210 may detect a visual representation of individual 2310 from the environment of user 100, such as a video of user 2310. As shown in
The tracked lip movement of individual 2310 may be used to separate if required, and selectively condition one or more sounds in the environment of user 100.
In addition to detecting images, apparatus 110 may be configured to detect one or more sounds in the environment of user 100. For example, microphone 1720 may detect one or more sounds 2421, 2422, and 2423 within environment 2400. In some embodiments, the sounds may represent voices of various individuals. For example, as shown in
Processor 210 may determine, based on lip movements and the detected sounds, which individuals in environment 2400 are speaking. For example, processor 210 may track lip movements associated with mouth 2311 to determine that individual 2310 is speaking. A comparative analysis may be performed between the detected lip movement and the received audio signals. In some embodiments, processor 210 may determine that individual 2310 is speaking based on a determination that mouth 2311 is moving at the same time as sound 2421 is detected. For example, when the lips of individual 2310 stop moving, this may correspond with a period of silence or reduced volume in the audio signal associated with sound 2421. In some embodiments, processor 210 may be configured to determine whether specific movements of mouth 2311 correspond to the received audio signal. For example, processor 210 may analyze the received audio signal to identify specific phonemes, phoneme combinations or words in the received audio signal. Processor 210 may recognize whether specific lip movements of mouth 2311 correspond to the identified words or phonemes. Various machine learning or deep learning techniques may be implemented to correlate the expected lip movements to the detected audio. For example, a training data set of known sounds and corresponding lip movements may be fed to a machine learning algorithm to develop a model for correlating detected sounds with expected lip movements. Other data associated with apparatus 110 may further be used in conjunction with the detected lip movement to determine and/or verify whether individual 2310 is speaking, such as a look direction of user 100 or individual 2310, a detected identity of user 2310, a recognized voiceprint of user 2310, etc.
Based on the detected lip movement, processor 210 may cause selective conditioning of audio associated with individual 2310. The conditioning may include amplifying audio signals determined to correspond to sound 2421 (which may correspond to a voice of individual 2310) relative to other audio signals. In some embodiments, amplification may be accomplished digitally, for example by processing audio signals associated with sound 2421 relative to other signals. Additionally, or alternatively, amplification may be accomplished by changing one or more parameters of microphone 1720 to focus on audio sounds associated with individual 2310. For example, microphone 1720 may be a directional microphone and processor 210 may perform an operation to focus microphone 1720 on sound 2421. Various other techniques for amplifying sound 2421 may be used, such as using a beamforming microphone array, acoustic telescope techniques, etc. The conditioned audio signal may be transmitted to hearing interface device 1710, and thus may provide user 100 with audio conditioned based on the individual who is speaking.
In some embodiments, selective conditioning may include attenuation or suppressing one or more audio signals not associated with individual 2310, such as sounds 2422 and 2423. Similar to amplification of sound 2421, attenuation of sounds may occur through processing audio signals, or by varying one or more parameters associated with microphone 1720 to direct focus away from sounds not associated with individual 2310.
In some embodiments, conditioning may further include changing a tone of one or more audio signals corresponding to sound 2421 to make the sound more perceptible to user 100. For example, user 100 may have lesser sensitivity to tones in a certain range and conditioning of the audio signals may adjust the pitch of sound 2421. For example, user 100 may experience hearing loss in frequencies above 10 kHz and processor 210 may remap higher frequencies (e.g., at 15 kHz) to 10 kHz. In some embodiments, processor 210 may be configured to change a rate of speech associated with one or more audio signals. Processor 210 may be configured to vary the rate of speech of individual 2310 to make the detected speech more perceptible to user 100. If speech recognition has been performed on the audio signal associated with sound 2421, conditioning may further include modifying the audio signal based on the detected speech. For example, processor 210 may introduce pauses or increase the duration of pauses between words and/or sentences, which may make the speech easier to understand. Various other processing may be performed, such as modifying the tone of sound 2421 to maintain the same pitch as the original audio signal, or to reduce noise within the audio signal.
The conditioned audio signal may then be transmitted to hearing interface device 1710 and then produced for user 100. Thus, in the conditioned audio signal, sound 2421 (may be louder and/or more easily distinguishable than sounds 2422 and 2423.
Processor 210 may be configured to selectively condition multiple audio signals based on which individuals associated with the audio signals are currently speaking. For example, individual 2310 and individual 2410 may be engaged in a conversation within environment 2400 and processor 210 may be configured to transition from conditioning of audio signals associated with sound 2421 to conditioning of audio signals associated with sound 2422 based on the respective lip movements of individuals 2310 and 2410. For example, lip movements of individual 2310 may indicate that individual 2310 has stopped speaking or lip movements associated with individual 2410 may indicate that individual 2410 has started speaking. Accordingly, processor 210 may transition between selectively conditioning audio signals associated with sound 2421 to audio signals associated with sound 2422. In some embodiments, processor 210 may be configured to process and/or condition both audio signals concurrently but only selectively transmit the conditioned audio to hearing interface device 1710 based on which individual is speaking. Where speech recognition is implemented, processor 210 may determine and/or anticipate a transition between speakers based on the context of the speech. For example, processor 210 may analyze audio signals associate with sound 2421 to determine that individual 2310 has reached the end of a sentence or has asked a question, which may indicate individual 2310 has finished or is about to finish speaking.
In some embodiments, processor 210 may be configured to select between multiple active speakers to selectively condition audio signals. For example, individuals 2310 and 2410 may both be speaking at the same time or their speech may overlap during a conversation. Processor 210 may selectively condition audio associated with one speaking individual relative to others. This may include giving priority to a speaker who has started but not finished a word or sentence or has not finished speaking altogether when the other speaker started speaking. This determination may also be driven by the context of the speech, as described above.
Various other factors may also be considered in selecting among active speakers. For example, a look direction of the user may be determined and the individual in the look direction of the user may be given higher priority among the active speakers. Priority may also be assigned based on the look direction of the speakers. For example, if individual 2310 is looking at user 100 and individual 2410 is looking elsewhere, audio signals associated with individual 2310 may be selectively conditioned. In some embodiments, priority may be assigned based on the relative behavior of other individuals in environment 2400. For example, if both individual 2310 and individual 2410 are speaking and more other individuals are looking at individual 2410 than individual 2310, audio signals associated with individual 2410 may be selectively conditioned over those associated with individual 2310. In embodiments where the identity of the individuals is determined, priority may be assigned based on the relative status of the speakers, as discussed previously in greater detail. User 100 may also provide input into which speakers are prioritized through predefined settings or by actively selecting which speaker to focus on.
Processor 210 may also assign priority based on how the representation of individual 2310 is detected. While individuals 2310 and 2410 are shown to be physically present in environment 2400, one or more individuals may be detected as visual representations of the individual (e.g., on a display device) as shown in
In step 2510, process 2500 may include receiving a plurality of images captured by a wearable camera from an environment of the user. The images may be captured by a wearable camera such as camera 1730 of apparatus 110. In step 2520, process 2500 may include identifying a representation of at least one individual in at least one of the plurality of images. The individual may be identified using various image detection algorithms, such as Haar cascade, histograms of oriented gradients (HOG), deep convolution neural networks (CNN), scale-invariant feature transform (SIFT), or the like. In some embodiments, processor 210 may be configured to detect visual representations of individuals, for example from a display device, as shown in
In step 2530, process 2500 may include identifying at least one lip movement or lip position associated with a mouth of the individual, based on analysis of the plurality of images. Processor 210 may be configured to identify one or more points associated with the mouth of the individual. In some embodiments, processor 210 may develop a contour associated with the mouth of the individual, which may define a boundary associated with the mouth or lips of the individual. The lips identified in the image may be tracked over multiple frames or images to identify the lip movement. Accordingly, processor 210 may use various video tracking algorithms, as described above.
In step 2540, process 2500 may include receiving audio signals representative of the sounds captured by a microphone from the environment of the user. For example, apparatus 110 may receive audio signals representative of sounds 2421, 2422, and 2423 captured by microphone 1720. In step 2550, process 2500 may include identifying, based on analysis of the sounds captured by the microphone, a first audio signal associated with a first voice and a second audio signal associated with a second voice different from the first voice. For example, processor 210 may identify an audio signal associated with sounds 2421 and 2422, representing the voice of individuals 2310 and 2410, respectively. Processor 210 may analyze the sounds received from microphone 1720 to separate the first and second voices using any currently known or future developed techniques or algorithms. Step 2550 may also include identifying additional sounds, such as sound 2423 which may include additional voices or background noise in the environment of the user. In some embodiments, processor 210 may perform further analysis on the first and second audio signals, for example, by determining the identity of individuals 2310 and 2410 using available voiceprints thereof. Alternatively, or additionally, processor 210 may use speech recognition tools or algorithms to recognize the speech of the individuals.
In step 2560, process 2500 may include causing selective conditioning of the first audio signal based on a determination that the first audio signal is associated with the identified lip movement associated with the mouth of the individual. Processor 210 may compare the identified lip movement with the first and second audio signals identified in step 2550. For example, processor 210 may compare the timing of the detected lip movements with the timing of the voice patterns in the audio signals. In embodiments where speech is detected, processor 210 may further compare specific lip movements to phonemes or other features detected in the audio signal, as described above. Accordingly, processor 210 may determine that the first audio signal is associated with the detected lip movements and is thus associated with an individual who is speaking.
Various forms of selective conditioning may be performed, as discussed above. In some embodiments, conditioning may include changing the tone or playback speed of an audio signal. For example, conditioning may include remapping the audio frequencies or changing a rate of speech associated with the audio signal. In some embodiments, the conditioning may include amplification of a first audio signal relative to other audio signals. Amplification may be performed by various means, such as operation of a directional microphone, varying one or more parameters associated with the microphone, or digitally processing the audio signals. The conditioning may include attenuating or suppressing one or more audio signals that are not associated with the detected lip movement. The attenuated audio signals may include audio signals associated with other sounds detected in the environment of the user, including other voices such as a second audio signal. For example, processor 210 may selectively attenuate the second audio signal based on a determination that the second audio signal is not associated with the identified lip movement associated with the mouth of the individual. In some embodiments, the processor may be configured to transition from conditioning of audio signals associated with a first individual to conditioning of audio signals associated with a second individual when identified lip movements of the first individual indicates that the first individual has finished a sentence or has finished speaking.
In step 2570, process 2500 may include causing transmission of the selectively conditioned first audio signal to a hearing interface device configured to provide sound to an ear of the user. The conditioned audio signal, for example, may be transmitted to hearing interface device 1710, which may provide sound corresponding to the first audio signal to user 100. Additional sounds such as the second audio signal may also be transmitted. For example, processor 210 may be configured to transmit audio signals corresponding to sounds 2421, 2422, and 2423. The first audio signal, which may be associated with the detected lip movement of individual 2310, may be amplified, however, in relation to sounds 2422 and 2423 as described above. In some embodiments, hearing interface 1710 device may include a speaker associated with an earpiece. For example, hearing interface device may be inserted at least partially into the ear of the user for providing audio to the user. Hearing interface device may also be external to the ear, such as a behind-the-ear hearing device, one or more headphones, a small portable speaker, or the like. In some embodiments, hearing interface device may include a bone conduction microphone, configured to provide an audio signal to user through vibrations of a bone of the user's head. Such devices may be placed in contact with the exterior of the user's skin, or may be implanted surgically and attached to the bone of the user.
Wearable Apparatus and Methods for Approving Transcription and/or Summary
A wearable apparatus consistent with the disclosed embodiments may be used in circumstances in which a user is using a device such as apparatus 110 in an environment of the user to capture images and/or record sound from the environment of the user. Apparatus 110 may perform additional actions, such as, recognizing people the user meets, separating recorded speech into segments in accordance with the different speakers, whether recognized by apparatus 110 or not, transcribing some or all of the recorded speech or the like.
As described above, audio signals captured from within the environment of a user may be processed prior to presenting the audio to the user. For example, in some situations, audio capture and processing capabilities may be used for purposes such as transcription of the spoken words, phrases, and sentences. In some embodiments, for example when the disclosed apparatus or methods are being used by a user, for example a physician, the apparatus may be used for capturing, transcribing and storing speech, for example by the user. In the example of the physician user, the physician may dictate the findings and recommended treatment, and the patient may receive referrals, prescriptions, further treatments, or the like based on the physician's input. Thus, it is highly important that the physician's summary and recommendations are transcribed and stored in an accurate manner.
To ensure accuracy of the transcription, the apparatus may enter a specific mode, for example, by recognizing that the user has said a predetermined phrase, such as “please transcribe,” “transcription please,” “start transcription,” or a similar phrase or command to begin transcribing the audio signals. While in this mode, the apparatus may transcribe certain audio signals, whether in real time as the audio is being captured, or retroactively for previously captured audio signals. After or during the transcription, for example after every sentence, after every pause, or after a predetermined time, the apparatus may transform the resulting text into speech, and read it to the user, for example via a speaker.
In some embodiments, the apparatus may enter a different mode, for example, a “summary mode.” While in this mode, the apparatus may summarize the transcribed text and the summary may be shown to the user. For example, a physician appointment may be summarized as “headaches, Tylenol® twice a day.” The transcribed text and/or summary may then be stored subject to the user's approval that it is correct. If incorrect, the user may amend the text, and the amended text may be stored. In some embodiments, the entire transcription, its summary, or both may be read out to the user and then stored subject to the user's approval or corrections. In some embodiments, the text or summary may be further analyzed, for example certain data items may be classified. Thus, the phrase “headaches, Tylenol® twice a day” may be structured into “user complaints: headaches; diagnosis: no pathological findings; treatment: Tylenol, dosage: twice a day.”
Thus, one aspect of the present disclosure relates to a wearable apparatus that may capture and store an audio signal and/or one or more images associated with an environment of a user. Another aspect of the disclosure may relate to a system that includes a wearable apparatus and a computing platform (e.g., computing device 120, and/or server 250, etc.) Yet another aspect of the disclosure may relate to methods of providing a transcription or summary based on the one or more audio signals and/or one or more images captured by the wearable apparatus.
In some embodiments, a system for processing audio signals is disclosed. For example, as illustrated in
In some embodiments, the system comprises a wearable device. In some embodiments, user 100 may wear a wearable apparatus (e.g., similar to the camera-based hearing aid device discussed above). Consistent with the disclosed embodiments, apparatus 110 may be positioned in various locations associated with user 100, as described previously. For example, apparatus 110 may be physically connected to a shirt, a necklace, a belt, glasses, a wrist strap, a button, etc.
In some embodiments, the wearable device comprises a microphone having an audio sensor configured to capture the audio signals from an environment of the user. As discussed above, apparatus 110 may include one or more microphones to receive one or more sounds associated with an environment of user 100. By way of example, apparatus 110 may comprise microphones 443, 444, as described with respect to
In some embodiments, the wearable device comprises a camera having an image sensor configured to capture at least one image from the environment of the user. For example, image sensor 220 may be part of a camera included in apparatus 110. It is contemplated that image sensor 220 may be associated with a variety of cameras, for example, a wide angle camera, a narrow angle camera, an IR camera, etc. In some embodiments, the camera may include a video camera. The one or more cameras (including one or more image sensors 220) may be configured to capture images from the surrounding environment (e.g., environment 2600) of user 100 and output an image signal. For example, the one or more cameras may be configured to capture individual still images or a series of images in the form of a video. The one or more cameras may be configured to generate and output one or more image signals representative of the one or more captured images. In some embodiments, the image signal may include a video signal. For example, when image sensor 220 is associated with a video camera, the video camera may output a video signal representative of a series of images captured as a video image by the video camera. By way of example, wearable device 110 worn by user 100 in
In some embodiments, the wearable apparatus comprises a processor. By way of example, as discussed above, apparatus 110 may include processor 210 (see
In some embodiments, the processor is programmed to receive the audio signals captured by the microphone. For example, one or more microphones 443, 444 associated with apparatus 110 may generate one or more audio signals, for example, audio signals 103, 2624, or other audios signals and transmit the audio signals to, for example, processor 210. Processor 210 may receive the one or more audio signals over a wired or wireless connection. Although only processor 210 is discussed here and in the following description, the described functions may be performed by any of processors 210, 210a, 210b (see
In some embodiments, the processor is programmed to receive the at least one image captured by the camera. By way of example, image sensor 220 associated with wearable apparatus 110 may capture an image from an environment 2600 of user 100. Image sensor 220 may be configured to transmit the one or more images captured by the camera to, for example, processor 210. Processor 210 may receive the one or more images over a wired or wireless connection.
In some embodiments, the processor is programmed to receive an instruction to enter into a transcription mode. In an example embodiment, user 100 may provide instructions and feedback to processor 210 using any suitable means (e.g., via audio signals or via a suitable interface). The interface may include keys, touch-sensitive areas, buttons, levers, knobs, or other input devices on apparatus 110. In some embodiments, the interface may be an application displayed, for example, on a secondary device coupled to the wearable device, such as a smartphone, a laptop, tablet, smart glasses, smartwatch, or another electronic device. User 100 may provide specific instructions to the wearable device (e.g., user 100 may request to enter a transcription mode). For example, user 100 may issue a spoken command such as “Transcription Mode,” “Enter Transcription Mode,” “Begin Transcription,” “Please Transcribe,” Transcription Please,” or “Start Transcription.” When the instructions are provided via voice commands from user 100, such instructions may be analyzed by an audio analysis application of apparatus 110 and then the disclosed system may be configured to generate a transcription as will be described below. In an example embodiment, the audio analysis application of apparatus 110 may be any suitable software application capable of transcribing human speech and executing instructions for entering into a transcription mode. In other embodiments, user 100 may use one or more of the input devices or an interface shown on a coupled device to enter a command to begin transcription.
In some embodiments, the processor is programmed to analyze the audio signals to generate a transcription. In some embodiments, the processor is programmed to generate the transcription from the audio signals after entering the transcription mode. As discussed, in some embodiments, the processor may be programmed to transcribe the audio signals after entering a transcription mode. In other embodiments, the processor may be programmed to transcribe the audio signals without specifically entering into a transcription mode. Rather, the processor may transcribe the audio signals at periodic intervals, after the audio signals are no longer being received, or when the processor detects a pause in speech in the audio signals.
In some embodiments, the processor may be programmed to detect voices or speech in the received audio signals. For example, processor 210 may use one or more artificial Intelligence (AI) engines, such as neural networks, trained on relevant training sets. In other examples, processor 210 may use one or more voice recognition algorithms, such as Hidden Markov Models, Dynamic Time Warping, neural networks, or other techniques to identify voices or speech in the received audio signals. Voice recognition component 2041 and/or processor 210 may access database 2050, which may include voiceprints of user 100 and/or one or more other individuals (e.g., individual 2610). Voice recognition component 2041 may analyze the audio signal to determine whether one or more segments of the audio signal match one or more voiceprints stored in database 2050. Accordingly, database 2050 may contain voiceprint data associated with a number of individuals (e.g., user 100, individual 2610). When processor 210 determines a match between, for example, one or more segments of the audio signal and one or more voiceprints stored in database 2050, processor 210 may identify the one or more segments of the audio signal as including speech. However, processor 210 may identify the one or more segments of the audio signal as including speech in other ways, and also if the audio signal does not include voice by an identified individual.
Having a speaker's voiceprint and a high-quality voiceprint may provide for fast and efficient way of identifying a voice or speech in the received audio signals. A high-quality voice print may be collected, for example, when user 100, or another individual (e.g., individual 2610, etc.) speaks alone, preferably in a quiet environment. By having a voiceprint of one or more speakers, it may be possible to separate voices in the audio signals almost in real time, e.g., with a minimal delay, using a sliding time window. The delay may be, for example 10 ms, 20 ms, 30 ms, 50 ms, 100 ms, or the like. Different time windows may be selected, depending on the quality of the voice print, on the quality of the captured audio, the difference in characteristics between speaker(s), the available processing resources, the required separation quality, or the like. In some embodiments, a voice print may be extracted from a segment of a conversation in which an individual (e.g., user 100, individual 2610) speaks alone, and then used for detecting the individual's voice in the received audio signals. However, a voice print, for example of user 100, may be a voice print generated upon a high quality audio signal which is pre-stored within the device.
In some embodiments, separating and/or detecting voices in the received audio signals may be performed as follows: spectral features, also referred to as spectral attributes, a spectral envelope, or a spectrogram may be extracted from a clean audio of a single speaker and fed into a pre-trained first neural network, which may generate or update a signature of the speaker's voice based on the extracted features. It will be appreciated that the voice signature may be generated using any other engine or algorithm and is not limited to a neural network. The audio may be for example, of one second of a clean voice. The output signature may be a vector representing the speaker's voice, such that the distance between the vector and another vector extracted from the voice of the same speaker is typically smaller than the distance between the vector and a vector extracted from the voice of another speaker. The speaker's model may be pre-generated from a captured audio. Alternatively or additionally, the model may be generated after a segment of the audio in which only the speaker speaks, followed by another segment in which the speaker and another speaker (or background noise) is heard, and which it is required to separate.
Then, to separate and/or identify speech in a noisy audio, a second pre-trained engine, such as a neural network, may receive the noisy audio and a speaker's signature and output an audio (which may also be represented as attributes) of the speech of the speaker as extracted from the noisy audio, separated from the other speech or background noise. It will be appreciated that the same or additional neural networks may be used to separate the voices of multiple speakers. For example, if there are two possible speakers, two neural networks may be activated, each with models of the same noisy output and one of the two speakers. Alternatively, a neural network may receive voice signatures of two or more speakers and output the voice of each of the speakers separately. Accordingly, the system may generate two or more different audio outputs, each comprising the speech of a respective speaker. In some embodiments, if separation is impossible, the input voice may only be cleaned from background noise.
In some embodiments, processor 210 may employ one or more trained machine learning models or neural networks to identify one or more segments of the audio signal received by processor 210 as comprising speech. For example, a set of training audio signals together with corresponding labels may be provided to train a machine learning model or a neutral network. One or more segments of the audio signal received by processor 210 may be presented as input to the trained machine learning model or neural network, which may output an indication regarding whether or not the one or more segments of the audio signal comprise speech.
Processor 210 may be programmed to transcribe the detected voices or speech in the audio signal using various speech-to-text algorithms or AI engines. Processor 210 may be programmed to execute one or more sound recognition modules in voice recognition component 2041 to transcribe the detected voices or speech in the received audio signals. In some embodiments, the one or more sound processing modules may allow processor 210 to convert one or more spoken words to text, using any known speech-to-text process or technology. It is contemplated that processor 210 may be programmed to store the transcribed text in, for example, database 2050 or in any other storage device associated with processor 210. In some embodiments, transcribing audio by a particular speaker may use a voice print or a specific model associated with the speaker.
In some embodiments, the processor is programmed to analyze the at least one image to generate the transcription. In some embodiments, the at least one processor may be programmed to detect voices or speech in the received audio signals and transcribe the voices or speech into the text using analysis of one or more images. In some embodiments, the processor is programmed to track lip movements represented in a plurality of images captured by the camera to generate the transcription. For example, processor 210 may analyze one or more images obtained by image sensor 220 to identify one or more mouth gestures for determining one or more spoken phonemes, syllables, or words. Processor 210 may be programmed to identify the one or more mouth gestures by, for example, tracking a movement of lips as described above. Processor 210 may use one or more video tracking algorithms, such as mean-shift tracking, contour tracking (e.g., a condensation algorithm), or various other techniques to track the movement of lips. Processor 210 may also be programmed to determine one or more mouth gestures based on a shape and/or variation of the shape of the lips over time. Processor 210 may be programmed to determine a spoken word based on the one or more mouth gestures based on one or more rules. For example, processor 210 may access one or more rules specifying one or more relationships between identified mouth gestures and spoken words, phonemes, or syllables. In some embodiments, lip tracking may be used in conjunction with or as an aid to audio processing, for example by identifying that when an “A” sound is heard, the mouth of one of the individuals is opened, thereby identifying the speaker which may lead to easier recognition afterwards. Processor 210 may be programmed to determine the spoken word or words for transcription based on the one or more rules, which may be stored, for example, in database 1950.
In some embodiments, processor 210 may be programmed to determine the one or more spoken words using a machine learning algorithm or neural network that may be trained using training examples. Examples of such algorithms may include support vector machines, Fisher's linear discriminant, nearest neighbor, k nearest neighbors, decision trees, random forests, neural networks of any type such as deep neural networks, or the like. By way of example, a set of training examples may include a plurality of images showing one or more mouth gestures associated with identified spoken phonemes, syllables, or words. For example, the training examples may include image, or image sequence samples including one or more words spoken by a plurality of speakers. The machine learning algorithm or neural network may be trained to identify one or more spoken phonemes, syllables, or words from captured images based on these and/or other training examples. The trained machine learning algorithm or neural network may be programmed to output text corresponding to the one or more spoken words when presented with one or more images of mouth gestures as inputs. In some embodiment, a trained neural network for identifying one or more words may be a separate and distinct neural network or may be an integral part of one or more other neural networks discussed above.
In some embodiments, the processor may be programmed to perform optical character recognition (OCR) on one or more images captured by the image sensor. For example, processor 210 may perform OCR on the one or more images captured by image sensor 220 and may identify textual matter in the one or more images based on the OCR result. Processor 210 may be programmed to transcribe the textual matter in the one or more images based on the OCR result.
In some embodiments, the processor is programmed to receive an instruction to enter into a summary mode. In an example embodiment, user 100 may provide feedback to processor 210 using any suitable means (e.g., via audio signals or via a suitable interface). The interface may include buttons, levers, knobs, touch screen of an associated device, or other input devices on apparatus 110. In some embodiments, the interface may be an application displayed, for example, on a device coupled to the wearable device, such as a smartphone, a laptop, tablet, smart glasses, smartwatch, or another electronic device. User 100 may provide specific instructions to the wearable device (e.g., user 100 may request to enter a summary mode). For example, user 100 may issue a spoken command such as “Please Summarize,” “Summary Mode,” “Enter Summary Mode,” “Begin Summary,” or “Start Summary.” When the instructions are provided via voice commands from user 100, such instructions may be analyzed by a language processing application of apparatus 110 and then the disclosed system may be programmed to generate a summary of the transcription as will be described below. In an example embodiment, the language processing application of apparatus 110 may be any suitable software application capable of transcribing human speech and executing instructions for entering into a transcription mode. In other embodiments, user 100 may use one or more of the input devices or an interface shown on a coupled device to enter a command to begin generating the summary of the transcription.
In some embodiments, the processor is programmed to generate a summary of the transcription. In some embodiments, the processor is programmed to generate the summary of the transcription after entering the summary mode. In other embodiments, the processor may be programmed to generate the summary without specifically entering into a summary mode. Rather, the processor may generate the summary at periodic intervals or after the transcription is complete.
In some embodiments, the at least one processor may be programmed to generate a summary based on the transcribed text. In some embodiments, the processor may be programmed to generate the summary by parsing the transcribed text, using at least one of syntactic parsing or semantic parsing. For example, processor 210 may analyze the transcript (transcribed text) of the audio signal and generate the summary by semantic parsing to convert the transcribed text into a machine understandable logical form. Processor 210 may parse the transcribed text to determine related concepts, subjects, contexts, or roles associated with some or all of the transcribed text. By way of example, when the transcribed text includes text such as “patient” or “age,” processor 210 may use semantic parsing to associate a context such as a name, gender, or age with the word patient based on the transcribed text and may generate a summary based on the associated context.
In some embodiments, processor 210 may additionally or alternatively analyze the transcribed text of the audio signal using syntactic parsing. For example, processor 210 may apply syntactic parsing to segment, tokenize, and tag the transcribed text and assign it to a tree structure that reveals the relationships between various tokens. Thus, for example, processor 210 may apply syntactic parsing to identify the subject and predicate of a sentence, a target associated with the sentence, and words or phrases that may modify the target. The identified subject, predicate, and target may be stored in the form of a tree structure that defines the relationships between these components. In one exemplary embodiment, processor 210 may employ a parser such as CoreNLP to perform syntactic parsing of the transcribed text. The concepts, subjects, contexts, or roles obtained by semantic parsing and/or the subject, predicate, target, and their relationships obtained by syntactic parsing may constitute the structured data generated by processor 210 from the transcribed text.
In some embodiments, processor 210 may employ one or more trained machine learning models or neural networks to identify the words or phrases to be included in the summary. For example, a set of training examples including transcribed text of audio signals together with corresponding summaries may be provided to train a machine learning model or a neutral network. Processor 210 may execute the trained machine learning model or neural network by providing transcribed text obtained from the received audio signals as input to the trained machine learning model or neural network, which may output a summary of the transcribed text.
In some embodiments, the processor is programmed to cause the summary to be displayed to the user. In some embodiments, the processor is programmed to cause the summary to be displayed to the user via a display associated with the wearable device. In some embodiments, the processor is programmed to cause the summary to be displayed to the user via a display of a secondary device associated with the wearable device. The summary may be displayed to a user on a device associated with the wearable device or on a secondary device being used by the user. In some embodiments, a display device such as a heads-up display, an augmented reality device, or a virtual reality device may be associated with the wearable device. By way of example, processor 210 of wearable device 110 may be programmed to transmit one or more signals via a wired or wireless connection to an associated display. The associated display in turn may be configured to receive the signals from the wearable device 110 and display the summary on the display. Additionally or alternatively, wearable device 110 may include a projector capable of projecting information on an external surface such as a screen or a wall in environment 2600 of the user. The projector associated with the display device may project the summary of the transcribed text onto an external surface such as a screen or a wall. In some embodiments, wearable device 110 may be associated with a secondary computing device (e.g., 120) that may include a display. Such secondary devices may include one or more of a smartphone, a laptop, a desktop computer, tablet, smartwatch, smart glasses, or any other type of computing device. The processor may be programmed to transmit signals to the associated secondary device via a wired or wireless connection. The secondary device in turn may be configured to receive the signals from the processor and display the summary of the transcribed text on a display associated with the secondary device. For example, processor 210 may be programmed to control feedback outputting unit 230 to transmit signals to a display device or a secondary device associated with the wearable device to cause a display of the summary.
In some embodiments, the processor is programmed to analyze at least one of the transcription or the summary and classify at least one item in the transcription or the summary into a least one category. As discussed above, the processor may analyze the transcribed text of the audio signals or the text in the generated summary using semantic or syntactic parsing. The processor may be programmed to identify relationships between words and one or more categories. In some embodiments, one or more categories may be predefined and may be stored. For example, one or more predefined categories may be stored in memory 550 and/or database 2050. In some embodiments, the disclosed system may allow a user of the wearable device to create the categories, either using voice commands or by providing inputs using one or more input devices (e.g., buttons, levers, knobs) or applications on one or more connected devices. Processor 210 may be programmed to use one or more algorithms as described above to identify one or more categories from a user's voice commands or from the inputs received via the one or more input devices or applications. Processor 210 may also be programmed to store the identified categories in, for example, memory 550 and/or database 2050. Further, processor 210 may be programmed to apply semantic and syntactic parsing on the transcribed text or the summary to assign one or more terms in the transcribed text or the summary to the one or more categories. By way of example, referring to the summary in
In some embodiments, the processor is programmed to receive a confirmation input from the user indicating that the displayed summary is correct. In an example embodiment, user 100 may provide feedback to processor 210 using any suitable means (e.g., via audio signals or via a suitable interface). The interface may include buttons, levers, knobs, touch areas, or other input devices on apparatus 110. User 100 may provide specific instructions to the wearable device (e.g., user 100 may specify that the displayed summary is correct). For example, user 100 may issue a spoken command such as “Summary Correct,” or “Accept Summary.” When the instructions are provided via voice commands from user 100, such instructions may be analyzed by a language processing application of apparatus 110 and then the disclosed system may be configured to accept the displayed summary without any changes. In other embodiments, user 100 may use one or more of the input devices discussed above or an interface shown on a coupled device to enter a confirmation input indicating that the summary is correct.
In some embodiments, the processor is programmed to cause a user interface component to be displayed to the user. As discussed above, the processor is programmed to display the summary of the transcribed text on a display device (e.g., display device associated with the wearable device or display of an associated secondary device). In some embodiments, the summary may be displayed in a graphical user interface (GUI) that may include one or more user interface components that may allow a user to provide inputs and/or to revise the summary. Such user interface components may include, for example, buttons, checkboxes, list boxes, pull down menus, scroll bars, one or more widgets to allow for editing of the displayed summary, and/or any other type of graphical user interface element that may allow the user to provide an input or revise the summary. For example, as illustrated in
In some embodiments, the processor is programmed to receive the confirmation input via the displayed user interface component. As discussed, the processor may be programmed to display the summary of the transcribed audio signals via a GUI that may include one or more input elements such as buttons, checkboxes, list boxes, or menus. A user may activate one or more of these elements using one or more input devices, for example, a keyboard, a mouse, a touch screen, a pointing device, and/or one or more input devices such as buttons, levers, and/or knobs associated with the wearable device. Activating the one or more GUI elements may include pressing, selecting, touching, or performing any other action, resulting in providing an input to the processor. In some embodiments, the user may provide an input by interacting with the one or more GUI elements in a virtual or augmented reality environment. The processor may be programmed to receive a confirmation indicating that the displayed summary is correct based on the received user input. By way of example,
In some embodiments, prior to receiving the confirmation input, the processor is programmed to receive from the user a correction to the summary. As discussed, the processor may be programmed to display the summary of the transcribed audio signals via a GUI that may include one or more input elements such as buttons, checkboxes, list boxes, or menus. Prior to providing an input (e.g., by pressing a button labeled “Accept”), a user may instead provide an input indicating that the user wishes to make a revision to the displayed summary. For example, as illustrated in
When the processor receives an input indicating that the summary is not correct, the processor may be programmed to display additional GUI elements on GUI 2710 to allow the user to make corrections. For example, processor 210 may cause display of one or more editing tools, and/or a virtual keyboard on GUI 2710 to allow a user to make edits and/or corrections to the displayed summary. In some embodiments, a user may additionally or alternatively use physical input devices, for example, such as keyboards, mice, joysticks, or digital pens to make additions, deletions, and/or alterations to the displayed summary. It is also contemplated that in some embodiments, the user may make corrections to the displayed summary using voice commands. As discussed above, the processor may use one or more voice recognition algorithms to determine the command issued by a user and make appropriate modifications (e.g., additions, deletions, substitutions) to the displayed summary.
In some embodiments, the processor is programmed to cause a corrected summary to be displayed to the user. The processor may be programmed to display the one or more corrections to the summary in real time as the user is making those corrections. Additionally or alternatively, the processor may be programmed to display a corrected version of the summary only after the user has completed making all the revisions. The processor may determine that the user has completed making all the revisions when there are no inputs detected from the user over a predetermined period of time. The processor may be programmed to display the corrected summary to the user in the same manner as discussed earlier. By way of example, processor 210 may cause the summary displayed on GUI 2710 to change as the user revises the summary. Alternatively, processor 210 may display an updated summary on GUI 2710, when processor 210 detects no inputs over a predetermined period of time.
In some embodiments, the processor is programmed to generate structured data from at least one of the transcription or the summary and display the structured data to the user. As discussed above, the processor may be programmed to generate structured data by classifying one or more words in the summary into one or more categories. The processor may use one or more techniques such as semantic or syntactic parsing to generate the structured data. In some embodiments, the processor may be programmed to use one or more trained machine learning models or neural networks to generate the structured data. For example, the one or more machine learning models or neural networks may be provided with training examples, including audio signals, transcribed text, or summaries of transcribed text together with a corresponding categorization into one or more categories may be provided as inputs. After the one or more machine learning models or neural networks have been trained, the processor may provide the audio, the transcription generated from the received audio signals, or a summary of the transcribed text as an input to the trained machine learning model or neural network. The trained machine learning model or neural network may in turn provide structured data by categorizing the summary text into one or more categories. In some embodiments, the processor may also be programmed to display the generated structured data in one or more formats, as will be described below, on a display associated with the wearable device or on a display of an associated secondary device.
In some embodiments, the processor is programmed to cause a plurality of structured display formats to be displayed to the user; receive a selection of a structured display format from the plurality of structured display formats; and display the structured data to the user in the selected structured display format. The processor may be programmed to display the structured data in many different formats. One such format discussed above, for example, with respect to
For example,
In some embodiments, the processor is programmed to cause the displayed summary to be stored. For example, on receiving an input confirming that the displayed summary is correct, the processor may be programmed to store the displayed summary in a local or remote storage. The displayed summary to be stored may be the summary initially generated by the processor from the transcription of the audio signals, and/or a summary corrected by the user as described above. For example, processor 210 may be programmed to store the summary as displayed on GUI 2760 in database 2050 or in any other storage device (e.g., memory 550, 550a, 550b) associated with processor 210. As another example, processor 210 may be programmed to store the summary displayed on GUI 2760 in
In some embodiments, the audio signals include first audio signals and the processor is programmed to transform the transcription into second audio signals representing speech. In some embodiments, the audio signals include first audio signals and the processor is programmed to transform the summary into second audio signals representing speech. For example, the first audio signals may be the one or more audio signals received by the processor from the one or more audio sensors associated with the wearable device. The processor may also be programmed to convert either the transcription of the first audio signals, a summary of the transcription, or both into second audio signals using a text-to-speech algorithms. Such algorithms may include one or more of WaveNet, DeepVoice, Tacatron, or any other algorithm that converts a textual input into an audio signal representing speech. By way of example, processor 210 may be programmed to execute one or more text-to-speech modules stored in database 2050 to convert the transcription and/or a summary of the transcription into audio signals representing speech.
In some embodiments, the processor is programmed to cause the second audio signals to be played via a speaker. In some embodiments, the speaker is a hearing aid, a speaker of the wearable device, or an external speaker associated with the wearable device. Feedback of information according to the disclosed embodiments may include audible feedback to user 100 (e.g., using a Bluetooth™ or other wired or wirelessly connected external speaker. For example, processor 210 may be programmed to transmit the second audio signals to feedback-outputting unit 230, which may include audio headphones, a hearing aid device, a speaker, a portable speaker carried or worn by user 100, a bone conduction headphone, or others. In some embodiments, processor 210 may transmit the second audio signals to an external feedback outputting unit 230 via a wireless transceiver 530, a wired connection, or some other communication interface. In some embodiments, feedback outputting unit 230 may be integrated into other devices, such as a Bluetooth™ headset of the user, glasses, a helmet (e.g., motorcycle helmets, bicycle helmets, etc.), a hat, etc. A user may be able to hear the speech associated with second audio signals using one or more of the devices described above.
In some embodiments, the processor is programmed to play the second audio signals at the end of a sentence, during a pause, or when a predetermined amount of time elapses after receiving the first audio signals. As discussed above the processor may be programmed to generate a transcription of the audio signals in real time or near real time. The processor may also be programmed to generate the second audio signals by performing text-to-speech conversion of the transcription in real time or near real time. In some embodiments, the processor may be programmed to play the second audio signals at the end of a sentence. For example, processor 210 may analyze the one or more audio signals 103 and/or 2624. Processor 210 may detect a period of silence in the audio signals, indicating an end of a sentence. Processor 210 may be programmed to determine the period of silence as one occurring after the end of a sentence by comparing the period to a first predetermined threshold amount of time. Processor 210 may be programmed to play the second audio signals (representing conversion of the transcription or summary into speech) during the period of silence. In some embodiments, processor may be programmed to play the second audio signals during a pause. Processor 210 may detect a period of silence in the audio signals, indicating a pause. Processor 210 may be programmed to determine the period of silence as a pause by comparing the period to a second predetermined threshold amount of time that may be longer than the first predetermined threshold of time. Processor 210 may be programmed to play the second audio signals (representing conversion of the transcription or summary into speech) after the second predetermined threshold amount of time. In some embodiments, processor 210 may be programmed to play the second audio signals when a predetermined amount of time elapses after receiving the first audio signals. This predetermined amount of time may be longer than the first or second predetermined threshold amounts of time. Processor 210 may determine that the first audio signals are no longer being received when processor 210 detects a period of silence greater than or equal to a third predetermined threshold amount of time that may be greater than the first and second predetermined threshold amounts of time. Processor 210 may be programmed to play the second audio signals (representing conversion of the transcription or summary into speech) after the third predetermined threshold amount of time has elapsed.
In some embodiments, the processor is a first processor and the system includes a second processor. In some embodiments, the microphone, the camera, and the first processor are included in the wearable device and the second processor is included in a secondary device, the secondary device including one of a tablet, a smartphone, a smartwatch, a laptop computer, or a desktop computer. For example, the disclosed system may include a wearable apparatus similar to apparatus 110 and a secondary device, for example, device 2660 (see
Device 2800 may include communication module 2816 configured to communicate with another device, for example, device 2660. Communication module 2816 may include one or more transceivers (e.g., wireless transceiver 530). It is contemplated that communication module 2816 may communicate with device 2660 via wired or wireless communication channels. Wireless transceiver 530 may use any known standard to transmit and/or receive data (e.g., Wi-Fi, Bluetooth®, Bluetooth Smart, 802.15.4, or ZigBee).
Device 2800 may also include storage 2818 that may be similar to memory 550 and/or database 2050. Storage 2818 may be configured to store fully or partially transcribed text generated by processor 210 from a received audio signal. Storage 2818 may also be configured to store a summary or structured data generated by processor 210 from the transcribed text.
Although voice capture device 2812 and speech recognition module 2814 have been described above as including instructions executable by processor 210, these components of device 2800 are not limited to software instructions. For example, one or more of voice capture device 2812 and speech recognition module 2814 may be implemented as a hardware module capable of performing the functions described above. By way of example, one or more of voice capture device 2812 and speech recognition module 2814 may be implemented as Application Specific Integrated Circuits (ASICs) or other electronic circuitry capable of performing the functions described above. Processor 210 may be programmed to cause the hardware and/or software associated with voice capture device 2812 and speech recognition module 2814 to perform their respective functions.
Device 2660 may be a device that may be used by, for example, user 100 to generate a transcription or a summary, to correct a transcription or a summary, to listen to a transcription or summary, and/or to review a transcription or summary. In some embodiments, device 2660 may include one of a tablet computer, a smartphone, a smartwatch, a laptop computer, or a desktop computer, etc. Device 2660 may include processor 540 and application 2830 for displaying and/or editing a transcription or summary of transcription. By way of example, application 2830 may be a program or application (e.g., an app) used by physicians or other care givers for handling patient data, including, for example, documenting a transcription or summary of a meeting with the patient. In some embodiments, application 2830 may include one or more instructions that when executed by a processor (e.g., processor 540) may cause a display device of device 2660 to display a document, display a GUI (e.g., GUI 2760) and/or one or more GUI elements (e.g., buttons 2764, 2766, 2768) for editing a transcription or summary of a transcription displayed on a display of device 2660.
Device 2660 may include communication module 2832, which may be similar to communication module 2816 of wearable device 2800. Communication module 2832 may include one or more transceivers (e.g., wireless transceiver 530). It is contemplated that communication module 2832 may communicate with wearable device 2800 via wired or wireless communication channels using any known standard to transmit and/or receive data (e.g., Wi-Fi, Bluetooth®, Bluetooth Smart, 802.15.4, or ZigBee). By way of example, communication module 2832 of device 2660 may receive audio signals and/or signals representing a transcription or a summary of a transcription from device 2660.
Device 2660 may include display module 2834 for displaying the transcription or summary of transcription. Display module 2834 may also be configured to display GUI 2760 comprising one or more GUI elements (e.g., buttons 2764, 2766, 2768) that may allow user 100 to confirm, revise, or display the transcription or summary of the transcription on a display of device 2660.
Device 2660 may include input module 2836 for receiving one or more user inputs for accepting the displayed transcription or summary of transcription, revising the displayed transcription or summary of transcription, or selecting a structured display format from among one or more available structured display formats displayed by display module 2834. Input module may include instructions, which when executed by a processor of device 2660 may allow a user of device 2660 (e.g., user 100) to perform the above-described functions. Device 2660 may also include storage 2838, which may be similar to storage 2818 of wearable device 2800. Device 2660 may also include speaker 2840, which may be similar to feedback outputting unit 230. In some embodiments, speaker 2840 may be configured to play audio signals corresponding to the transcription and/or the summary of the transcription to user 100. It is also contemplated that in some embodiments, speaker 2840 may be configured to play the audio signals generated by, for example, voice capture device 2812. For example, processor 210 of wearable device 2800 may be configured to transmit the audio signals generated by voice capture device 2812 to processor 540 of secondary device 2660 via communication modules 2816 and 2832. Processor 540 of secondary device 2660 may be programmed to play the audio signals received from processor 210.
It is contemplated that one or more of communication module 2832, display module 2834, or user input module 2836 may be implemented as part of application 2830. It is also contemplated that in some embodiments, one or more of voice capture device 2812 and/or speech recognition module 2814 may additionally or alternatively be implemented on device 2660.
Although application 2830, display module 2834, and/or input module 2836 have been described above as including instructions executable by a processor of device 2660, these components of device 2660 are not limited to software instructions. For example, one or more of application 2830, display module 2834, and/or input module 2836 may be implemented as hardware modules capable of performing the functions described above. By way of example, one or more of application 2830, display module 2834, and/or input module 2836 may be implemented as an ASIC or via other electronic circuitry. A processor associated with device 2660 may be programmed to cause the hardware and/or software associated with application 2830, display module 2834, and/or input module 2836 to perform their respective functions.
The audio signals received by wearable device 2800, transcriptions of the audio signals, and/or summaries of the audio signals may be stored on either device (e.g., in storage 2818 of wearable device 2800 or storage 2838 of secondary device 2660) or on a remote device for example on cloud storage. Processing of one or more instructions, for example, to convert audio signals into transcribed text, generate a summary of the transcription, receive inputs and/or any other processing tasks may be distributed in any manner between the processors of 2660 and 2800. For example some of the functions described above may be performed by processor 210 and the same of other remaining functions may be performed by processor 540 and vice-versa.
In some embodiments, the first processor is programmed to receive the audio signals and the at least one image; analyze the audio signals to generate the transcription; generate the summary of the transcription; and transmit the summary to the second processor. For example, as discussed above wearable apparatus 110 may include first processor 210 that may perform one or more of these functions. In particular, first processor 210 of wearable apparatus 110 (or wearable device 2800) may receive the audio signals and/or one or more images, analyze the audio signals to generate a transcription, and generate the summary of the transcription as described earlier. First processor 210 may also be programmed to transmit the generated transcription and/or the summary via, for example, communications module 2816 to second processor 540 of secondary device 2660. In some embodiments, the first processor is programmed to receive the audio signals and the at least one image; and transmit the summary to the second processor, while the second processor is programmed to analyze the audio signals to generate the transcription; generate the summary of the transcription. For example, as described above, first processor 210 of wearable apparatus 110 (or wearable device 2800) may receive the audio signals and/or one or more images and may be programmed to transmit the received audio signals and/or one or more images via, for example, communications module 2816 to second processor 540 of secondary device 2660. As will be described below, processor 540 of secondary device 2660 may be programmed to analyze the audio signals to generate a transcription, and generate the summary of the transcription.
In some embodiments, the second processor is programmed to cause the summary to be displayed to a user; receive the confirmation input from the user; and cause the displayed summary to be stored. By way of example, secondary device 2660 may include processor 540, which may be the second processor of the disclosed system. Processor 540 may execute instructions associated with application 2830 to display the transcription and/or the summary received from first processor 210. Processor 540 may execute, for example, display module 2834 to display the transcription and/or summary on a display associated with secondary device 2660. Processor 540 may also display a GUI (e.g., GUI 2760) along with GUI elements, for example, buttons 2764, 2766, and/or 2768 in a manner as discussed earlier. Processor 540 may also be programmed to receive a signal in response to user selection of buttons 2764, 2766, and/or 2768. For example, when user 100 presses the “Accept” button (e.g., button 2764), processor 540 may receive a signal indicating that the user has confirmed that the summary is correct. In response, processor 540 may store the summary in a memory (e.g., memory 550b) associated with processor 540 and/or in a database (e.g., database 2050). In some embodiments, processor 540 may transmit the summary to processor 210 of wearable device 110 (or 2800) and processor 210 may execute text storage module 2818 to store the summary in a memory (e.g., memory 550a) associated with processor 210 and/or in a database (e.g., 2050). However, the processing may be distributed in other manners between the processors. For example, the second processor may perform part or all of analyzing the audio signals to generate the transcription and generating the summary of the transcription. In some embodiments, the second processor is also programmed to analyze the audio signals to generate the transcription; generate the summary of the transcription. For example, processor 540 of secondary device 2660 may receive the audio signals and/or one or more images from first processor 210, analyze the audio signals to generate a transcription, and generate the summary of the transcription as described earlier.
In step 2902, process 2900 includes receiving audio signals representative of the sounds captured by a microphone from an environment of a user. For example, microphones 443, 444, and/or 1720 may capture one or more sounds from an environment (e.g., 2600) of user 100. As discussed above, processor 210 may be programmed to receive an audio signal that may include one or more of audio signals 103, 2624, etc., associated with user 100, individual 2610, and/or other individuals, respectively.
In step 2904, process 2900 includes receiving at least one image captured by an image sensor of a wearable apparatus. For example, the one or more images may be captured by a wearable camera such as a camera including image sensor 220 of apparatus 110 from an environment (e.g., 2600) of user 100.
In step 2906, process 2900 includes generating a transcription based on the audio signals and/or the at least one image. For example, processor 210 may be programmed to transcribe the speech included in the audio signals generated by microphones 443, 444, and/or 1720. Processor 210 may be programmed to transcribe the speech associated with a particular speaker (e.g., user 100, individual 2610, etc.) or the speech of all the speakers. Processor 210 may be programmed to transcribe the audio signal (e.g., 103 and/or 2624) using various speech-to-text algorithms. Processor 210 may be programmed to execute one or more sound recognition modules in voice recognition component 2041 to transcribe some or all of the audio signal (e.g., 103 and/or 2624). In some embodiments, the one or more sound processing modules may allow processor 210 to convert one or more spoken words to text, using any known speech-to-text process or technology. In some embodiments, processor 210 may first separate the captured audio according to the speaker in each segment, and transcribe the audio of each speaker separately. Separating the speaker and transcribing audio may or may not create, retrieve or use voice prints of the speakers.
In some embodiments, processor 210 may analyze one or more images obtained by image sensor 220 to identify one or more mouth gestures for determining one or more spoken words. Processor 210 may be programmed to identify the one or more mouth gestures by, for example, tracking a movement of lips of individual 2610 in the one or more images obtained by image sensor 220. Processor 210 may also be programmed to determine one or more mouth gestures based on a shape and/or variation of the shape of the lips over time. Processor 210 may be programmed to determine a spoken word based on the one or more mouth gestures based on one or more rules and/or using a trained neural network or machine learning model as described above. In some embodiments, processor 210 may perform optical character recognition on the one or more images obtained by image sensor 220 for generating the transcription.
In step 2908, process 2900 includes generating a summary and/or structured information from the transcription. For example, processor 210 may analyze the transcript (transcribed text) of the audio signal and generate the summary using one or more semantic and/or syntactic parsing techniques. Processor 210 may parse the transcribed text to determine related concepts, subjects, contexts, or roles associated with some of all of the transcribed text. Processor 210 may generate the summary based on the concepts, subjects, contexts, or roles. In some embodiments, processor 210 may also generate structure data from the transcribed text.
In step 2910, process 2900 includes displaying text or playing audio of the transcription and/or summary. For example, processor 210 may cause a display associated with wearable apparatus 110 (e.g., heads up display, augmented reality device, or virtual reality device) or a display associated with a secondary device (e.g., desktop or laptop computer, smartphone, smartwatch, tablet) to display the transcription or a summary of the transcription. In some embodiments, processor 210 may be programmed to display the summary in the form or one or more structured display formats (e.g., tabular format). In some embodiments, processor 210 may additionally or alternatively convert the transcription or the summary of the transcription into audio signals using one or more text-to-speech algorithms such as WaveNet, DeepVoice, Tacatron, or any other algorithm that converts a textual input into an audio signal representing speech. Processor 210 may transmit the audio signals to a hearing aid or an external speaker, which in turn may play the audio signals representing the transcription and/or the summary.
In step 2912, process 2900 includes receiving corrections to the transcription and/or the summary of the transcription. For example, user 100 may provide feedback to processor 210 using any suitable means (e.g., via audio signals or via a suitable interface). The interface may include buttons, levers, knobs, or other input devices on apparatus 110. User 100 may provide specific instructions to the wearable device (e.g., user 100 may specify that the displayed summary is correct). When the instructions are provided via voice commands from user 100, such instructions processor 210 may analyze the voice commands using a language processing application stored in, for example, database 2050 to identify the command. For example, processor 210 may determine from the voice command that user 100 wishes to revise or correct the summary. In other embodiments, user 100 may use one or more of the input devices discussed above or an interface shown on a coupled device to provide an input, indicating that user 100 wishes to revise or correct the summary. For example, as illustrated in
When the processor receives an input indicating that the summary is not correct, the processor may be programmed to display additional GUI elements on GUI 2710 to allow the user to make corrections. For example, processor 210 may cause display of one or more editing tools, and/or a virtual keyboard on GUI 2710 to allow user 100 to make edits and/or corrections to the displayed summary. In some embodiments, user 100 may additionally or alternatively use physical input devices, for example, such as keyboards, mice, joysticks, or digital pens to make additions, deletions, and/or alterations to the displayed summary. It is also contemplated that in some embodiments, user 100 may make corrections to the displayed summary using voice commands. As discussed above, the processor may use one or more voice recognition algorithms to determine the command issued by a user and make appropriate modifications (e.g., additions, deletions, substitutions) to the displayed summary.
In step 2914, process 2900 includes receiving confirmation input that the transcription and/or the summary is correct. For example, user 100 may provide feedback to processor 210 using any suitable means (e.g., via audio signals or via a suitable interface). The interface may include buttons, levers, knobs, or other input devices on apparatus 110. User 100 may provide specific instructions to the wearable device (e.g., user 100 may specify that the displayed summary is correct). For example, user 100 may issue a spoken command such as “Summary Correct,” or “Accept Summary.” When the instructions are provided via voice commands from user 100, such instructions may be analyzed by a language processing application of apparatus 110 and then the disclosed system may be configured to accept the displayed summary without any changes. In other embodiments, user 100 may use one or more of the input devices discussed above or an interface shown on a coupled device to enter a confirmation input indicating that the summary is correct.
In step 2916, process 2900 includes storing the transcription and/or the summary. For example, upon receiving an input confirming that the displayed summary is correct, the processor may be programmed to store the displayed summary in a local or remote storage. The displayed summary to be stored may be the summary initially generated by the processor from the transcription of the audio signals, or alternatively may be a summary corrected by the user as described above. For example, processor 210 may be programmed to store the summary as displayed on GUI 2760 in database 2050 or in any other storage device (e.g., memory 550, 550a, 550b) associated with processor 210. As another example, processor 210 may be programmed to store the summary displayed on GUI 2760 in
It will be appreciated that the foregoing description may be implemented on devices other than hearing aids, such as any device and in particular wearable device designed for capturing audio and images of a surrounding of a person. The foregoing description has been presented for purposes of illustration. It is not exhaustive and is not limited to the precise forms or embodiments disclosed. Modifications and adaptations will be apparent to those skilled in the art from consideration of the specification and practice of the disclosed embodiments. Additionally, although aspects of the disclosed embodiments are described as being stored in memory, one skilled in the art will appreciate that these aspects can also be stored on other types of computer readable media, such as secondary storage devices, for example, hard disks or CD ROM, or other forms of RAM or ROM, USB media, DVD, Blu-ray, Ultra HD Blu-ray, or other optical drive media.
Computer programs based on the written description and disclosed methods are within the skill of an experienced developer. The various programs or program modules can be created using any of the techniques known to one skilled in the art or can be designed in connection with existing software. For example, program sections or program modules can be designed in or by means of .Net Framework, .Net Compact Framework (and related languages, such as Visual Basic, C, etc.), Java, C++, Objective-C, HTML, HTML/AJAX combinations, XML, or HTML with included Java applets.
Moreover, while illustrative embodiments have been described herein, the scope of any and all embodiments having equivalent elements, modifications, omissions, combinations (e.g., of aspects across various embodiments), adaptations and/or alterations as would be appreciated by those skilled in the art based on the present disclosure. The limitations in the claims are to be interpreted broadly based on the language employed in the claims and not limited to examples described in the present specification or during the prosecution of the application. The examples are to be construed as non-exclusive. Furthermore, the steps of the disclosed methods may be modified in any manner, including by reordering steps and/or inserting or deleting steps. It is intended, therefore, that the specification and examples be considered as illustrative only, with a true scope and spirit being indicated by the following claims and their full scope of equivalents.
This application claims the benefit of priority of U.S. Provisional Patent Application No. 63/229,515, filed on Aug. 5, 2021, the contents of which are incorporated herein by reference in their entirety.
Number | Date | Country | |
---|---|---|---|
63229515 | Aug 2021 | US |