Patent Application
20220358929
The present application claims the priority of Chinese Patent Application No. 202110502922.2, filed on May 8, 2021, with the title of “VOICE ACTIVITY DETECTION METHOD AND APPARATUS, ELECTRONIC DEVICE AND STORAGE MEDIUM.” The disclosure of the above application is incorporated herein by reference in its entirety.
The present disclosure relates to the field of artificial intelligence technologies, and particularly to a voice activity detection method and apparatus, an electronic device and a storage medium in the fields of deep learning, intelligent voices, or the like.
In recent years, applications of human-machine interaction products are more and more popular; for example, the human-machine interaction products may be often found in places, such as bank halls, department stores, hospitals, or the like.
In order to perform an accurate voice interaction, voice activity detection (VAD) is usually required to be performed on collected voice data, and accuracy of a voice activity detection result directly affects a success rate of the human-machine interaction, or the like.
Currently, a voice detection model obtained by a training operation is usually used to perform the voice activity detection (i.e., detection of a voice start point and a voice end point) on the voice data. However, under a complex noise environment, especially when noise is ambient voices (background voices), this method usually has a poor detection result; that is, the accuracy of the detection result is not high.
The present disclosure provides a voice activity detection method and apparatus, an electronic device and a storage medium.
A method for voice activity detection includes acquiring time-aligned voice data and video data; performing a first detection of a voice start point and a voice end point of the voice data using a voice detection model obtained by a training operation; performing a second detection of a lip movement start point and a lip movement end point of the video data; and correcting a result of the first detection using a result of the second detection, and taking a corrected result as a voice activity detection result.
An electronic device includes at least one processor; and a memory communicatively connected with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a method for voice activity detection, wherein the method includes: acquiring time-aligned voice data and video data; performing a first detection of a voice start point and a voice end point of the voice data using a voice detection model obtained by a training operation; performing a second detection of a lip movement start point and a lip movement end point of the video data; and correcting a result of the first detection using a result of the second detection, and taking a corrected result as a voice activity detection result.
There is provided a non-transitory computer readable storage medium with computer instructions stored thereon, wherein the computer instructions are used for causing a computer to perform a method for voice activity detection, wherein the method includes: acquiring time-aligned voice data and video data; performing a first detection of a voice start point and a voice end point of the voice data using a voice detection model obtained by a training operation; performing a second detection of a lip movement start point and a lip movement end point of the video data; and correcting a result of the first detection using a result of the second detection, and taking a corrected result as a voice activity detection result.
An embodiment in the above-mentioned disclosure has the following advantages or beneficial effects. The voice detection model and a lip movement detection technology are combined, and the voice start point and the voice end point detected by the voice detection model are corrected using the lip movement detection result, thereby improving accuracy of the voice activity detection result, or the like.
It should be understood that the statements in this section are not intended to identify key or critical features of the embodiments of the present disclosure, nor limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.
The drawings are used for better understanding the present solution and do not constitute a limitation of the present disclosure. In the drawings,
The following part will illustrate exemplary embodiments of the present disclosure with reference to the drawings, including various details of the embodiments of the present disclosure for a better understanding. The embodiments should be regarded only as exemplary ones. Therefore, those skilled in the art should appreciate that various changes or modifications can be made with respect to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for clarity and conciseness, the descriptions of the known functions and structures are omitted in the descriptions below.
In addition, it should be understood that the term “and/or” only describes an association relationship between associated objects, and indicates that three relationships may exist. For example, A and/or B may indicate three cases: only A exists; both A and B exist; and only B exists. In addition, in this specification, the symbol “/” generally indicates that associated objects have a relationship of “or”.
step 101: acquiring time-aligned voice data and video data;
step 102: performing a first detection of a voice start point (vad_begin) and a voice end point (vad_end) of the voice data using a voice detection model obtained by a training operation;
step 103: performing a second detection of a lip movement start point (lip_begin) and a lip movement end point (lip_end) of the video data; and
step 104: correcting a result of the first detection using a result of the second detection result, and taking a corrected result as a voice activity detection result.
It is observed that, in the solution of the above-mentioned method embodiment, the voice detection model and a lip movement detection technology are combined, and the voice start point and the voice end point detected by the voice detection model are corrected using the lip movement detection result, thereby improving accuracy of the voice activity detection result, or the like.
In practical applications, a human-machine interaction product with a photographic apparatus (for example, a camera) may be used to collect the voice data and the video data for a same scenario; for example, when a user a performs a voice conversation with the human-machine interaction product, the voice data and the video data of the user a may be collected respectively.
The collected voice data and video data may be aligned in time. The voice data and the video data may be processed differently; the first detection of the voice start point and the voice end point of the voice data may be performed using the voice detection model obtained by the training operation, and the second detection of the lip movement start point and the lip movement end point of the video data may be performed.
A way of performing the first detection is a prior art.
As shown in
As shown in
It is assumed that noise exists at a head of the voice data and is an ambient voice, and
Therefore, the lip movement detection technology is introduced in the solution of the present disclosure, and besides the first detection of the voice start point and the voice end point of the voice data utilizing the voice detection model, the second detection of the lip movement start point and the lip movement end point of the video data is performed.
In an embodiment of the present disclosure, the second detection may be performed on the video data using a lip movement detection model obtained by a training operation, so as to obtain the lip movement start point and the lip movement end point of the face in a video.
There is no limitation on a way of performing the training operation to obtain the lip movement detection model. For example, a face picture in each picture frame in the video data may be captured using a face detection technology, or the like; generally speaking, each picture frame only includes one face picture, lip points of the face lips in each face picture may be extracted, each face picture may be labeled, the face picture with the moving lips is labeled with 1, the face picture without the moving lips is labeled with 0, and then, the lip movement detection model may be trained using the lip points of each face picture and a corresponding labeling result.
During the second detection, lip points may be obtained in the same way and input into the lip movement detection model, so as to obtain the output detection result of 1 or 0 which indicates whether the face lips in the face picture move or not.
Generally speaking, at least 5 picture frames are required to determine the lip movement start point, which is similar to the processing way in the first detection, and also has a purpose of preventing the misjudgment; that is, a lip movement lasting for a period of time is considered to be a real lip movement (i.e., a lip movement caused by speaking), thus preventing a detection error caused by a short-time lip movement due to other reasons.
Therefore, the time (T_lip_begin_detected) when the lip movement start point is detected has a certain delay compared with the lip movement start point, and with the aforementioned 5 frames as an example, if a frame rate is 20, the delay is 250 ms, and the lip movement start point may be obtained by a backing operation.
In this way, the lip movement start point and the lip movement end point may be accurately and efficiently detected, thereby laying a good foundation for a subsequent processing operation, or the like.
The obtained first detection result may be corrected using the obtained second detection result.
To this end, several states, such as a voice detection state and a lip movement detection state, may be predefined, the voice detection state may further include a voiced state (state_vad_speech) and a voiceless state (state_vad_no_speech), and the lip movement detection state may further include a lip movement state (state_lip) and a lip-movement-free state (state_no_lip).
The states are as follows:
A. voiced state: a state in a time after the voice start point is detected and before the corresponding voice end point is detected;
B. voiceless state: a state in a time other than the voiced state, i.e., a state in a time before the voice start point is detected and after the voice end point is detected;
C. lip movement state: a state in a time after the lip movement start point is detected and before the corresponding lip movement end point is detected; and
D. lip-movement-free state: a state in a time other than the lip movement state, i.e., a state in a time before the lip movement start point is detected and after the lip movement end point is detected.
In an embodiment of the present disclosure, when the first detection result is corrected using the second detection result, the following processing methods may be adopted:
1) when the voice detection state is the voiced state and the lip movement detection state is the lip-movement-free state, if the lip movement start point is detected and meets a predefined time requirement, the detected lip movement start point may be used as a determined voice end point and a new voice start point; and
2) when the voice detection state is the voiced state and the lip movement detection state is the lip movement state, if the lip movement end point is detected, the detected lip movement end point may be used as the determined voice end point and the new voice start point.
In the method 1), when the voice detection state is the voiced state and the lip movement detection state is the lip-movement-free state, if the lip movement start point is detected, whether the lip movement start point meets the predefined time requirement may be further determined; in an embodiment of the present disclosure, a difference between the time when the lip movement start point is detected and the time when the voice start point (i.e., the voice start point corresponding to the current voiced state) is last detected is greater than a predefined threshold may be determined, if yes, the lip movement start point may be determined to meet the predefined time requirement, and correspondingly, the detected lip movement start point may be used as the determined voice end point and the new voice start point.
That is, forced segmentation may be performed, and the detected lip movement start point is used as the voice end point corresponding to the last detected voice start point and the new voice start point.
Generally speaking, the voice data between the last detected voice start point and the corresponding voice end point obtained after the processing operation using the above-mentioned method is the noise occurring before the voice for the human-machine interaction, which is equivalent to segmentation of the noise and the voice for the human-machine interaction, and the voice start point is re-determined, thereby improving the accuracy of the detection result of the voice start point, or the like.
In addition, since the lip movement start point is detected, the lip movement detection state is switched from the lip-movement-free state to the lip movement state.
In the method 2), when the voice detection state is the voiced state and the lip movement detection state is the lip movement state, if the lip movement end point is detected, the detected lip movement end point may be used as the determined voice end point and the new voice start point.
That is, forced segmentation may be performed, and the detected lip movement end point is used as the voice end point corresponding to the newly obtained voice start point and the new voice start point.
Generally speaking, the voice data between the newest voice start point and the corresponding voice end point obtained after the processing operation using the above-mentioned method is the voice for the human-machine interaction, which is equivalent to the segmentation of the noise and the voice for the human-machine interaction, and the voice end point is re-determined, thereby improving the accuracy of the detection result of the voice end point, or the like.
Using the processing operations in the methods 1) and 2), the detected voice start point and voice end point are corrected by means of the lip movement detection, such that the head noise and end noise may be removed, thereby improving the accuracy of the voice activity detection result, or the like.
In the solution of the present disclosure, the existence of the second detection does not affect normal execution of the first detection; that is, the detection of the voice start point and the voice end point may be performed according to the existing method, and the voice detection state may be determined according to the detected voice start point and voice end point, and only from the perspective of external output, the first detection result may be corrected using the second detection result.
In an embodiment of the present disclosure, if the face lips in the video are determined not to be occluded, the second detection of the lip movement start point and the lip movement end point of the video data may be performed.
That is, when the face lips in the video are determined to be occluded, for example, due to wear of a mask, the second detection may not be performed; that is, only the first detection is performed, and the first detection result is used as the voice activity detection result, thereby well incorporating the existing implementation, flexibly coping with various possible situations, and guaranteeing a sequential proceeding effect of the man-machine interaction, or the like.
In combination with the above introduction,
It should be noted that for simplicity of description, the above-mentioned embodiment of the method is described as combinations of a series of acts, but those skilled in the art should understand that the present disclosure is not limited by the described order of acts, as some steps may be performed in other orders or simultaneously according to the present disclosure. Further, those skilled in the art should also understand that the embodiments described in this specification are preferred embodiments and that acts and modules referred to are not necessary for the present disclosure.
The above is a description of an embodiment of the method, and an embodiment of an apparatus according to the present disclosure will be further described below.
The acquired voice data and video data may be processed differently; the first detecting module 702 may perform the first detection of the voice start point and the voice end point of the voice data using the voice detection model obtained by the training operation, and the second detecting module 703 may perform the second detection of the lip movement start point and the lip movement end point of the video data. That is, the lip movement detection technology is introduced, and besides the first detection of the voice start point and the voice end point of the voice data utilizing the voice detection model, the second detection of the lip movement start point and the lip movement end point of the video data is performed.
In an embodiment of the present disclosure, the second detecting module 703 may perform the second detection on the video data using a lip movement detection model obtained by a training operation, so as to obtain the lip movement start point and the lip movement end point of the face in a video.
There is no limitation on a way of performing the training operation to obtain the lip movement detection model. For example, a face picture in each picture frame in the video data may be captured using a face detection technology, or the like; generally speaking, each picture frame only includes one face picture, lip points of the face lips in each face picture may be extracted, each face picture may be labeled, the face picture with the moving lips is labeled with 1, the face picture without the moving lips is labeled with 0, and then, the lip movement detection model may be trained using the lip points of each face picture and a corresponding labeling result.
During the second detection, lip points may be obtained in the same way and input into the lip movement detection model, so as to obtain the output detection result of 1 or 0 which indicates whether the face lips in the face picture move or not.
Further, the correcting module 704 may correct the first detection result using the second detection result to obtain the desired voice activity detection result.
To this end, several states, such as a voice detection state and a lip movement detection state, may be predefined, the voice detection state may further include a voiced state and a voiceless state, and the lip movement detection state may further include a lip movement state and a lip-movement-free state.
The voiced state is a state in a time after the voice start point is detected and before the corresponding voice end point is detected; the voiceless state is a state in a time other than the voiced state, i.e., a state in a time before the voice start point is detected and after the voice end point is detected; the lip movement state is a state in a time after the lip movement start point is detected and before the corresponding lip movement end point is detected; and the lip-movement-free state is a state in a time other than the lip movement state, i.e., a state in a time before the lip movement start point is detected and after the lip movement end point is detected.
Correspondingly, in an embodiment of the present disclosure, when the voice detection state is the voiced state and the lip movement detection state is the lip-movement-free state, if the lip movement start point is detected and meets a predefined time requirement, the correcting module 704 may use the detected lip movement start point as a determined voice end point and a new voice start point.
In an embodiment of the present disclosure, the meeting of the predefined time requirement may mean that a difference between the time when the lip movement start point is detected and the time when the voice start point is last detected is greater than a predefined threshold.
That is, when the voice detection state is the voiced state and the lip movement detection state is the lip-movement-free state, if the lip movement start point is detected, the difference between the time when the lip movement start point is detected and the time when the voice start point (i.e., the voice start point corresponding to the current voiced state) is last detected is greater than the predefined threshold may be further determined, if yes, the lip movement start point may be determined to meet the predefined time requirement, and correspondingly, the detected lip movement start point may be used as the determined voice end point and the new voice start point. That is, forced segmentation may be performed, and the detected lip movement start point is used as the voice end point corresponding to the last detected voice start point and the new voice start point.
In an embodiment of the present disclosure, when the voice detection state is the voiced state and the lip movement detection state is the lip movement state, if the lip movement end point is detected, the correcting module 704 may further use the detected lip movement end point as the determined voice end point and the new voice start point.
That is, forced segmentation may be performed, and the detected lip movement end point is used as the voice end point corresponding to the newly obtained voice start point and the new voice start point.
In an embodiment of the present disclosure, the second detecting module 703 may further perform the second detection on the video data when the face lips in the video are determined not to be occluded. That is, when the face lips in the video are determined to be occluded, for example, due to wear of a mask, the second detection may not be performed; that is, only the first detection is performed, and the first detection result is used as the voice activity detection result,
For the specific work flow of the embodiment of the apparatus shown in
In conclusion, in the solution of the apparatus according to the embodiment of the present disclosure, the voice detection model and a lip movement detection technology are combined, and the voice start point and the voice end point detected by the voice detection model are corrected using the lip movement detection result, thereby improving accuracy of the voice activity detection result, or the like.
The solution of the present disclosure may be applied to the field of artificial intelligence, and particularly relates to the fields of deep learning, intelligent voices, or the like.
Artificial intelligence is a subject of researching how to cause a computer to simulate certain thought processes and intelligent behaviors (for example, learning, inferring, thinking, planning, or the like) of a human, and includes both hardware-level technologies and software-level technologies. Generally, the hardware technologies of the artificial intelligence include technologies, such as a sensor, a dedicated artificial intelligence chip, cloud computing, distributed storage, big data processing, or the like; the software technologies of the artificial intelligence mainly include a computer vision technology, a voice recognition technology, a natural language processing technology, a machine learning/deep learning technology, a big data processing technology, a knowledge graph technology, or the like.
According to the embodiment of the present disclosure, there are also provided an electronic device, a readable storage medium and a computer program product.
As shown in
The plural components in the device 800 are connected to the I/O interface 805, and include: an input unit 806, such as a keyboard, a mouse, or the like; an output unit 807, such as various types of displays, speakers, or the like; the storage unit 808, such as a magnetic disk, an optical disk, or the like; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, or the like. The communication unit 809 allows the device 800 to exchange information/data with other devices through a computer network, such as the Internet, and/or various telecommunication networks.
The computing unit 801 may be a variety of general and/or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphic processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, or the like. The computing unit 801 performs the methods and processing operations described above, such as the method according to the present disclosure. For example, in some embodiments, the method according to the present disclosure may be implemented as a computer software program tangibly contained in a machine readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program may be loaded and/or installed into the device 800 via the ROM 802 and/or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the method according to the present disclosure may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the method according to the present disclosure by any other suitable means (for example, by means of firmware).
Various implementations of the systems and technologies described herein above may be implemented in digital electronic circuitry, integrated circuitry, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems on chips (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and/or combinations thereof. The systems and technologies may be implemented in one or more computer programs which are executable and/or interpretable on a programmable system including at least one programmable processor, and the programmable processor may be special or general, and may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input apparatus, and at least one output apparatus.
Program codes for implementing the method according to the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or a controller of a general purpose computer, a special purpose computer, or other programmable data processing apparatuses, such that the program code, when executed by the processor or the controller, causes functions/operations specified in the flowchart and/or the block diagram to be implemented. The program code may be executed entirely on a machine, partly on a machine, partly on a machine as a stand-alone software package and partly on a remote machine, or entirely on a remote machine or a server.
In the context of the present disclosure, the machine readable medium may be a tangible medium which may contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine readable medium may be a machine readable signal medium or a machine readable storage medium. The machine readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine readable storage medium may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read only memory (ROM), an erasable programmable read only memory (EPROM or flash memory), an optical fiber, a portable compact disc read only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide interaction with a user, the systems and technologies described here may be implemented on a computer having: a display apparatus (for example, a cathode ray tube (CRT) or liquid crystal display (LCD) monitor) for displaying information to a user; and a keyboard and a pointing apparatus (for example, a mouse or a trackball) by which a user may provide input for the computer. Other kinds of apparatuses may also be used to provide interaction with a user; for example, feedback provided for a user may be any form of sensory feedback (for example, visual feedback, auditory feedback, or tactile feedback); and input from a user may be received in any form (including acoustic, speech or tactile input).
The systems and technologies described here may be implemented in a computing system (for example, as a data server) which includes a back-end component, or a computing system (for example, an application server) which includes a middleware component, or a computing system (for example, a user computer having a graphical user interface or a web browser through which a user may interact with an implementation of the systems and technologies described here) which includes a front-end component, or a computing system which includes any combination of such back-end, middleware, or front-end components. The components of the system may be interconnected through any form or medium of digital data communication (for example, a communication network). Examples of the communication network include: a local area network (LAN), a wide area network (WAN) and the Internet.
A computer system may include a client and a server. Generally, the client and the server are remote from each other and interact through the communication network. The relationship between the client and the server is generated by virtue of computer programs which run on respective computers and have a client-server relationship to each other. The server may be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so as to overcome the defects of high management difficulty and weak service expansibility in conventional physical host and virtual private server (VPS) service. The server may also be a server of a distributed system, or a server incorporating a blockchain. The cloud computing technology is a technical system in which an elastically extensible shared physical or virtual resource pool is accessed through a network, resources may include servers, operating systems, networks, software, applications, storage devices, or the like, and the resources may be deployed and managed in a self-service mode according to needs; the cloud computing technology may provide an efficient and powerful data processing capacity for technical applications and model training of artificial intelligence, blockchains, or the like.
It should be understood that various forms of the flows shown above may be used and reordered, and steps may be added or deleted. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, which is not limited herein as long as the desired results of the technical solution disclosed in the present disclosure may be achieved.
The above-mentioned implementations are not intended to limit the scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made, depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure all should be included in the extent of protection of the present disclosure.
| Number | Date | Country | Kind |
|---|---|---|---|
| 202110502922.2 | May 2021 | CN | national |
