Determining audience state or interest using passive sensor data

Abstract

This document describes techniques and apparatuses for determining an audience's state or interest using passive sensor data. The techniques receive sensor data that measures an audience during presentation of a media program. The techniques, based on this sensor data, then determine a state or interest of the audience members during the media program.

Description

RELATED APPLICATION

This application claims priority to U.S. patent application Ser. No. 14/815,321, filed on Jul. 31, 2015, which, in turn, claims priority to U.S. patent application Ser. No. 13/316,351, filed on Dec. 9, 2011 the disclosure of which is incorporated in its entirety by reference herein.

BACKGROUND

Advertisers and media providers want to know how many people are watching a particular television show or other media program. Current approaches often compile hand-written logs recorded by a small fraction of the potential viewing public. Using these logs and statistical modeling, current approaches estimate a total number of viewers based on a number of the small fraction that logged that they viewed the program. By so doing, advertisers and media providers may agree to advertising costs for commercials played during the television show, as well as learn what kinds of shows people wish to watch.

SUMMARY

This document describes techniques and apparatuses for determining an audience's state or interest using passive sensor data. The techniques receive sensor data that measures an audience during presentation of a media program. This sensor data can be passively obtained, and thus not annoy audience members or introduce inaccuracies common when relying on members to record their attendance or reaction to a media program. The techniques, based on this sensor data, then determine a state or interest of the audience members during the media program. The techniques may determine multiple states of an audience member over the course of the media program, such as a member laughing, then clapping, and then looking away from the media program. The techniques may also or instead determine an audience member's interest in the media program, such as a viewer having a high interest by laughing during a comedy or staring quietly during a drama. Further, the techniques may determine viewers' states and interest at various points in a media program effective to enable analysis of portions of a media program.

This summary is provided to introduce simplified concepts for determining audience state or interest using passive sensor data, which is further described below in the Detailed Description. This summary is not intended to identify essential features of the claimed subject matter, nor is it intended for use in determining the scope of the claimed subject matter.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of techniques and apparatuses for determining audience state or interest using passive sensor data are described with reference to the following drawings. The same numbers are used throughout the drawings to reference like features and components:

FIG. 1 illustrates an example environment in which techniques for determining audience state or interest using passive sensor data can be implemented.

FIG. 2 is an illustration of an example computing device that is local to the audience of FIG. 1.

FIG. 3 is an illustration of an example remote computing device that is remote to the audience of FIG. 1, as well as a network through which the remote computing device communicates with the computing device of FIG. 2.

FIG. 4 illustrates example methods for determining audience state or interest using passive sensor data.

FIG. 5 illustrates an interest graph having interest levels for one person over forty time periods during presentation of a media program.

FIG. 6 illustrates an example device in which techniques for determining audience state or interest using passive sensor data can be implemented.

DETAILED DESCRIPTION

Overview

This document describes techniques and apparatuses for determining audience state or interest using passive sensor data. By so doing, the techniques can determine not only that a viewer watched a media program, but what portions the viewer watched and how intently the viewer watched those portions, as well as a wealth of other valuable information.

Consider, for example, a 30-minute situational comedy, such as The Office, which is typically 22 minutes in total content with eight minutes of advertisements interspersed. Media providers, media creators, and advertisers would like to know how many people watched the program but also how viewers reacted to various parts of the program and the interspersed advertisements. For example, did many viewers laugh at a particular joke? Did they leave the room when a character in the show got into an embarrassing situation? Did they leave or stay for an advertisement? Did they watch an advertisement with interest (e.g., because they found it funny) or stay but ignore it?

The described techniques and apparatuses can provide answers to these questions by determining an audience's state or interest. Assume, for the above example, that two people are watching The Office in their home. The techniques can determine that the first person was present for all 30 minutes but did not pay attention to 7 of 16 total advertisements, laughed at minute 4, 13, and 19 of the 22 total minutes for the show, looked away and talked during minutes 15 and 16, and paid close attention and then cheered for an advertisement for a new adventure movie.

Similarly, the techniques can determine that the second person was not in the room until minute 3, left at minutes 12-15, was looking away and talking during minute 16, laughed at minutes 19 and 20, left for 9 of the 16 advertisements, and paid close attention to the new adventure movie advertisement and an advertisement for a women's clothing sale.

Based on this information, media providers, media creators, and advertisers can more-accurately price advertisements, determine future content likely to be well received (e.g., 80% of viewers laughed at the joke at minute 19, but only 10% laughed at the situation at minute 7, so future programs should include similar jokes but not similar situations), and determine future content of advertisements (e.g., the clothing-sale advertisement was well received by women in the audience, so structure future ads similarly).

This is but one example of how techniques and/or apparatuses for determining an audience's state or interest using passive sensor data can be performed. Techniques and/or apparatuses that determine an audience's state or interest using passive sensor data are referred to herein separately or in conjunction as the “techniques” as permitted by the context. This document now turns to an example environment in which the techniques can be embodied, after which various example methods for performing the techniques are described.

Example Environment

FIG. 1 is an illustration of an example environment 100 in which the techniques may determine an audience's state or interest using passive sensor data. Environment 100 includes a media presentation device 102, an audience-sensing device 104, a state module 106, and an interest module 108.

Media presentation device 102 presents a media program to an audience 110 having one or more persons 112. A media program can include, alone or in combination, a television show, a movie, a music video, a video clip, an advertisement, a blog, a web page, an e-book, a computer game, a song, a tweet, or other audio and/or video media. Audience 110 can include one or more multiple persons 112 that are in locations enabling consumption of a media program presented by media presentation device 102 and measurement by audience-sensing device 104. In audience 110 three persons are shown: 112-1, 112-2, and 112-3.

Audience-sensing device 104 is capable of passively sensing audience 110 and providing sensor data for audience 110 to state module 106 and/or interest module 108 (sensor data shown provided at arrow 114). In this context, sensor data is passive by not requiring active participation of persons in the measurement of those persons. Examples of active sensor data include data recorded by persons in an audience, such as with hand-written logs, active entry of a user's impressions through selection by the user of buttons on a remote control, and data sensed from users through biometric sensors worn by persons in the audience. Passive sensor data can include data sensed using emitted light or other signals sent by audience-sensing device 104, such as with an infrared sensor bouncing emitted infrared light off of persons or the audience space (e.g., a couch, walls, etc.) and sensing the light that returns. Examples of passive sensor data and ways in which it is measured are provided in greater detail below.

Audience-sensing device 104 may or may not process sensor data prior to providing it to state module 106 and/or interest module 108. Thus, sensor data may be or include raw data or processed data, such as: RGB (Red, Green, Blue) frames; infrared data frames; depth data; heart rate; respiration rate; a person's head orientation or movement (e.g., coordinates in three dimensions, x, y, z, and three angles, pitch, tilt, and yaw); facial (e.g., gaze of eyes, eyebrow, eyelid, nose, and mouth) orientation, movement, or occlusion; skeleton's orientation, movement, or occlusion; audio, which may include information indicating orientation sufficient to determine from which person the audio originated or directly indicating which person, or what words were said, if any; thermal readings sufficient to determine or indicating presence and locations of one of persons 112; and distance from the audience-sensing device 104 or media presentation device 102. In some cases audience-sensing device 104 includes infrared sensors (e.g., webcams, Kinect cameras), stereo microphones or directed audio microphones, eye-tracking sensors, and a thermal reader (in addition to infrared sensors), though other sensing apparatuses may also or instead be used.

State module 106 receives sensor data and determines, based on the sensor data, states of persons 112 in audience 110 (shown at arrow 116). States include, for example: sad, talking, disgusted, afraid, smiling, scowling, placid, surprised, angry, laughing, screaming, clapping, waving, cheering, looking away, looking toward, leaning away, leaning toward, asleep, or departed, to name just a few.

The talking state can be a general state indicating that a person is talking, though it may also include subcategories based on the content of the speech, such as talking about the media program (related talking) or talking that is unrelated to the media program (unrelated talking). State module 106 can determine which talking category through speech recognition.

State module 106 may also or instead determine, based on sensor data, a number of persons, a person's identity and/or demographic data (arrow 118), or engagement (arrow 120) during presentation. Identity indicates a unique identity for one of persons 112 in audience 110, such as Susan Brown. Demographic data classifies one of persons 112, such as 5 feet, 4 inches tall, young child, and male or female. Engagement indicates whether a person is likely to be paying attention to the media program, such as based on that person's presence or facial orientation. Engagement, in some cases, can be determined by state module 106 with lower-resolution or less-processed sensor data compared to that used to determine states. Even so, engagement can be useful in measuring an audience, whether on its own or to determine a person's interest using interest module 108.

Interest module 108 determines, based on sensor data (arrow 114) and/or a person's engagement or state (shown with dashed-line arrow 122) and information about the media program (shown at media type arrow 124), that person's interest level (arrow 126) in the media program. Interest module 108 may determine, for example, that multiple laughing states for a media program intended to be a serious drama indicate a low level of interest and conversely, that for a media program intended to be a comedy, that multiple laughing states indicate a high level of interest.

State module 106 and interest module 108 can be local to audience 110, and thus media presentation device 102 and audience-sensing device 104, though this is not required. An example embodiment where state module 106 and interest module 108 are local to audience 110 is shown in FIG. 2. In some cases, however, state module 106 and/or interest module 108 are remote from audience 110, which is illustrated in FIG. 3.

FIG. 2 is an illustration of an example computing device 202 that is local to audience 110. Computing device 202 includes or has access to media presentation device 102, audience-sensing device 104, one or more processors 204, and computer-readable storage media (“media”) 206. Media 206 includes an operating system 208, state module 106, interest module 108, media program(s) 210, each of which may include or have associated program information 212. Note that in this illustrated example, media presentation device 102, audience-sensing device 104, state module 106, and interest module 108 are included within a single computing device, such as a desktop computer having a display, forward-facing camera, microphones, audio output, and the like. Each of these entities 102-108, however, may be separate from or integral with each other in one or multiple computing devices or otherwise. As will be described in part below, media presentation device 102 can be integral with audience-sensing device 104 but be separate from state module 106 or interest module 108.

As shown in FIG. 2, computing device(s) 202 can each be one or a combination of various devices, here illustrated with six examples: a laptop computer 202-1, a tablet computer 202-2, a smart phone 202-3, a set-top box 202-4, a desktop 202-5, and a gaming system 202-6, though other computing devices and systems, such as televisions with computing capabilities, netbooks, and cellular phones, may also be used. Note that three of these computing devices 202 include media presentation device 102 and audience-sensing device 104 (laptop computer 202-1, tablet computer 202-2, smart phone 202-3). One device excludes—but is in communication with—media presentation device 102 and audience-sensing device 104 (desktop 202-5). Two others exclude media presentation device 102 and may or may not include audience-sensing device 104, such as in cases where audience-sensing device 104 is included within media presentation device 102 (set-top box 202-4 and gaming system 202-6).

FIG. 3 is an illustration of an example remote computing device 302 that is remote to audience 110. FIG. 3 also illustrates a communications network 304 through which remote computing device 302 communicates with audience-sensing device 104 (not shown, but embodied within, or in communication with, computing device 202). Communication network 304 may be the Internet, a local-area network, a wide-area network, a wireless network, a USB hub, a computer bus, another mobile communications network, or a combination of these.

Remote computing device 302 includes one or more processors 306 and remote computer-readable storage media (“remote media”) 308. Remote media 308 includes state module 106, interest module 108, and media program(s) 210, each of which may include or have associated program information 212. Note that in this illustrated example, media presentation device 102 and audience-sensing device 104 are physically separate from state module 106 and interest module 108, with the first two local to an audience viewing a media program and the second two operating remotely. Thus, as will be described in greater detail below, sensor data is passed from audience-sensing device 104 to one or both of state module 106 or interest module 108, which can be communicated locally (FIG. 2) or remotely (FIG. 3).

These and other capabilities, as well as ways in which entities of FIGS. 1-3 act and interact, are set forth in greater detail below. These entities may be further divided, combined, and so on. The environment 100 of FIG. 1 and the detailed illustrations of FIGS. 2 and 3 illustrate some of many possible environments capable of employing the described techniques.

Example Methods

FIG. 4 depicts methods 400 for determining an audience's state or interest using passive sensor data. The methods of FIG. 4 are shown as sets of blocks that specify operations performed but are not necessarily limited to the order shown for performing the operations by the respective blocks. In portions of the following discussion reference may be made to environment 100 of FIG. 1 and entities detailed in FIGS. 2-3, reference to which is made for example only. The techniques are not limited to performance by one entity or multiple entities operating on one device.

Block 402 senses or receives sensor data for an audience or person, the sensor data passively sensed during presentation of a media program to the audience or person.

Consider, for example, a case where an audience includes three persons 112, persons 112-1, 112-2, and 112-3 all of FIG. 1. Assume that media presentation device 102 includes an LCD display and speakers through which the media program is rendered, and is in communication with set-top box 202-4 of FIG. 2. Here audience-sensing device 104 is a forward-facing high-resolution red-green-blue sensor, an infrared sensor, and two microphones capable of sensing sound and location, which is integral with set-top box 202-4 or media presentation device 102. Assume also that the media program 210 being presented is a PG-rated animated movie named Incredible Family, which is streamed from a remote source and through set-top box 202-4. Set-top box 202-4 presents Incredible Family with six advertisements, spaced one at the beginning of the movie, three in a three-ad block, and two in a two-ad block.

Sensor data is received for all three persons 112 in audience 110; for this example consider first person 112-1. Assume here that, over the course of Incredible Family, that audience-sensing device 104 measures, and then provides at block 402, the following at various times for person 112-1:

• • Time 1, head orientation 3 degrees, no or low-amplitude audio.
  • Time 2, head orientation 24 degrees, no audio.
  • Time 3, skeletal movement (arms), high-amplitude audio.
  • Time 4, skeletal movement (arms and body), high-amplitude audio.
  • Time 5, head movement, facial-feature changes, moderate-amplitude audio.
  • Time 6, detailed facial orientation data, no audio.
  • Time 7, skeletal orientation (missing), no audio.
  • Time 8, facial orientation, respiration rate.

Block 404 determines, based on the sensor data, a state of the person during the media program. In some cases block 404 determines a probability for the state or multiple probabilities for multiple states, respectively. For example, block 404 may determine a state likely to be correct but with less than full certainty (e.g., 90% chance that the person is laughing). Block 404 may also or instead determine that multiple states are possible based on the sensor data, such as a sad or placid state, and probabilities for each (e.g., sad state 65%, placid state 35%).

Block 404 may also or instead determine demographics, identity, and/or engagement. Further, methods 400 may skip block 404 and proceed directly to block 406, as described later below.

In the ongoing example, state module 106 receives the above-listed sensor data and determines the following corresponding states for person 112-1:

• • Time 1: Looking toward.
  • Time 2: Looking away.
  • Time 3: Clapping.
  • Time 4: Cheering.
  • Time 5: Laughing.
  • Time 6: Smiling.
  • Time 7: Departed.
  • Time 8: Asleep.

At Time 1 state module 106 determines, based on the sensor data indicating a 3-degree deviation of person 112-1's head from looking directly at the LCD display and a rule indicating that the looking toward state applies for deviations of less than 20 degrees (by way of example only), that person 112-1's state is looking toward. Similarly, at Time 2, state module 106 determines person 112-1 to be looking away due to the deviation being greater than 20 degrees.

At Time 3, state module 106 determines, based on sensor data indicating that person 112-1 has skeletal movement in his arms and audio that is high amplitude that person 112-1 is clapping. State module 106 may differentiate between clapping and other states, such as cheering, based on the type of arm movement (not indicated above for brevity). Similarly, at Time 4, state module 106 determines that person 112-1 is cheering due to arm movement and high-amplitude audio attributable to person 112-1.

At Time 5, state module 106 determines, based on sensor data indicating that person 112-1 has head movement, facial-feature changes, and moderate-amplitude audio, that person 112-1 is laughing. Various sensor data can be used to differentiate different states, such as screaming, based on the audio being moderate-amplitude rather than high-amplitude and the facial-feature changes, such as an opening of the mouth and a rising of both eyebrows.

For Time 6, audio-sensing device 104 processes raw sensor data to provide processed sensor data, and in this case facial recognition processing to provide detailed facial orientation data. In conjunction with no audio, state module 106 determines that the detailed facial orientation data (here upturned lip corners, amount of eyelids covering eyes) that person 112-1 is smiling.

At Time 7, state module 106 determines, based on sensor data indicating that person 112-1 has skeletal movement moving away from the audience-sensing device 104, that person 112-1 is departed. The sensor data may indicate this directly as well, such as in cases where audience-sensing device 104 does not sense person 112-1's presence, either through no skeletal or head readings or a thermal signature no longer being received.

At Time 8, state module 106 determines, based on sensor data indicating that person 112-1's facial orientation has not changed over a certain period (e.g., eyes have not blinked) and a steady, slow respiration rate that person 112-1 is asleep.

These eight sensor readings are simplified examples for purpose of explanation. Sensor data may include extensive data as noted elsewhere herein. Further, sensor data may be received measuring an audience every fraction of a second, thereby providing detailed data for tens, hundreds, and thousands of periods during presentation of a media program and from which states may be determined.

Returning to methods 400, block 404 may determine demographics, identity, and engagement in addition to a person's state. State module 106 may determine or receive sensor data from which to determine demographics and identity or receive, from audience-sensing device 104, the demographics or identity. Continuing the ongoing example, the sensor data for person 112-1 may indicate that person 112-1 is John Brown, that person 112-2 is Lydia Brown, and that person 112-3 is Susan Brown, for example. Or sensor data may indicate that person 112-1 is six feet, four inches tall and male (based on skeletal orientation), for example. The sensor data may be received with or include information indicating portions of the sensor data attributable separately to each person in the audience. In this present example, however, assume that audience-sensing device 104 provides three sets of sensor data, with each set indicating the identity of the person along with the sensor data.

Also at block 404, the techniques may determine an engagement of an audience or person in the audience. As noted, this determination can be less refined than that of states of a person, but nonetheless is useful. Assume for the above example, that sensor data is received for person 112-2 (Lydia Brown), and that this sensor data includes only head and skeletal orientation:

• • Time 1, head orientation 0 degrees, skeletal orientation upper torso forward of lower torso.
  • Time 2, head orientation 2 degrees, skeletal orientation upper torso forward of lower torso.
  • Time 3, head orientation 5 degrees, skeletal orientation upper torso approximately even with lower torso.
  • Time 4, head orientation 2 degrees, skeletal orientation upper torso back from lower torso.
  • Time 5, head orientation 16 degrees, skeletal orientation upper torso back from lower torso.
  • Time 6, head orientation 37 degrees, skeletal orientation upper torso back from lower torso.
  • Time 7, head orientation 5 degrees, skeletal orientation upper torso forward of lower torso.
  • Time 8, head orientation 1 degree, skeletal orientation upper torso forward of lower torso.

State module 106 receives this sensor data and determines the following corresponding engagement for Lydia Brown:

• • Time 1: Engagement High.
  • Time 2: Engagement High.
  • Time 3: Engagement Medium-High.
  • Time 4: Engagement Medium.
  • Time 5: Engagement Medium-Low.
  • Time 6: Engagement Low.
  • Time 7: Engagement High.
  • Time 8: Engagement High.

At Times 1, 2, 7, and 8, state module 106 determines, based on the sensor data indicating a 5-degree-or-less deviation of person 112-2's head from looking directly at the LCD display and skeletal orientation of upper torso forward of lower torso (indicating that Lydia is leaning forward to the media presentation) that Lydia is highly engaged in Incredible Family at these times.

At Time 3, state module 106 determines that Lydia's engagement level has fallen due to Lydia no longer leaning forward. At Time 4, state module 106 determines that Lydia's engagement has fallen further to medium based on Lydia leaning back, even though she is still looking almost directly at Incredible Family.

At Times 5 and 6, state module 106 determines Lydia is less engaged, falling to Medium-Low and then Low engagement based on Lydia still leaning back and looking slightly away (16 degrees) and then significantly away (37 degrees), respectively. Note that at Time 7 Lydia quickly returns to a High engagement, which media creators are likely interested in, as it indicates content found to be exciting or otherwise captivating.

Methods 400 may proceed directly from block 402 to block 406, or from block 404 to block 406 or block 408. If proceeding to block 406 from block 404, the techniques determine an interest level based on the type of media being presented and the person's engagement or state. If proceeding to block 406 from block 402, the techniques determine an interest level based on the type of media being presented and the person's sensor data, without necessarily first or independently determining the person's engagement or state.

Continuing the above examples for persons 112-1 and 112-2, assume that block 406 receives states determined by state module 106 at block 404 for person 112-1 (John Brown). Based on the states for John Brown and information about the media program, interest module 108 determines an interest level, either overall or over time, for Incredible Family. Assume here that Incredible Family is both an adventure and a comedy program, with portions of the movie marked as having either of these media types. While simplified, assume that Times 1 and 2 are marked as comedy, Times 3 and 4 are marked as adventure, Times 5 and 6 are marked as comedy, and that Times 7 and 8 are marked as adventure. Revisiting the states determined by state module 106, consider the following again:

• • Time 1: Looking toward.
  • Time 2: Looking away.
  • Time 3: Clapping.
  • Time 4: Cheering.
  • Time 5: Laughing.
  • Time 6: Smiling.
  • Time 7: Departed.
  • Time 8: Asleep.

Based on these states, state module 106 determines for Time 1 that John Brown has a medium-low interest in the content at Time 1—if this were of an adventure or drama type, state module 106 may determine John Brown to instead be highly interested. Here, however, due to the content being comedy and thus intended to elicit laughter or a similar state, interest module 108 determines that John Brown has a medium-low interest at Time 1. Similarly, for Time 2, interest module 108 determines that John Brown has a low interest at Time 2 because his state is not only not laughing or smiling but is looking away.

At Times 3 and 4, interest module 108 determines, based on the adventure type for these times and states of clapping and cheering, that John Brown has a high interest level. At time 6, based on the comedy type and John Brown smiling, that he has a medium interest at this time.

At Times 7 and 8, interest module 108 determines that John Brown has a very low interest. Here the media type is adventure, though in this case interest module 108 would determine John Brown's interest level to be very low for most types of content.

As can be readily seen, advertisers, media providers, and media creators can benefit from knowing a person's interest level. Here assume that the interest level is provided over time for Incredible Family, along with demographic information about John Brown. With this information from numerous demographically similar persons, a media creator may learn that male adults are interested in some of the adventure content but that most of the comedy portions are not interesting.

Consider, by way of a more-detailed example, FIG. 5, which illustrates an interest graph 500 having interest levels 502 for forty time periods 504 over a portion of a media program. Here assume that the media program is a movie that includes other media programs—advertisements—at time periods 18 to 30. Interest module 108 determines, as shown, that the person begins with a medium interest level, and then bounces between medium and medium-high, high, and very high interest levels to time period 18. During the first advertisement, which covers time periods 18-22, interest module 108 determines that the person has a medium low interest level. For time periods 23 to 28, however, interest module 108 determines that the person has a very low interest level (because he is looking away and talking or left the room, for example). For the last advertisement, which covers time period 28 to 32, however, interest module 108 determines that the person has a medium interest level for time periods 29 to 32—most of the advertisement. This can be valuable information the person stayed for the first advertisement, left for the middle advertisement and the beginning of the last advertisement, and returned, with medium interest, for most of the last advertisement. Contrast this resolution and accuracy of interest with some conventional approaches, which likely would provide no information about how many of the people that watched the movie actually watched the advertisements, which ones, and with what amount of interest. If this is a common trend with the viewing public, prices for advertisements in the middle of a block would go down, and other advertisement prices would be adjusted as well. Or, advertisers and media providers might learn to play shorter advertisement blocks having only two advertisements, for example. Interest levels 502 also provide valuable information about portions of the movie itself, such as through the very high interest level at time period 7 and the waning interest at time periods 35-38.

Note that, in some cases, engagement levels, while useful, may be less useful or accurate than states and interest levels. For example, state module 106 may determine, for just engagement levels, that a person is not engaged if the person's face is occluded (blocked) and thus not looking at the media program. If the person's face is blocked by that person's hands (skeletal orientation) and audio indicates high-volume audio, state module 106, when determining states, may determine the person to be screaming. A screaming state indicates, in conjunction with the content being horror or suspense, an interest level that is very high. This is but one example of where an interest level can be markedly different from that of an engagement level.

As noted above, methods 400 may proceed directly from block 402 to block 406. In such a case, interest module 108, either alone or in conjunction with state module 106, determines an interest level based on the type of media (including multiple media types for different portions of a media program) and the sensor data. By way of example, interest module 108 may determine that for sensor data for John Brown at Time 4, which indicates skeletal movement (arms and body), and high-amplitude audio, and a comedy, athletics, conflict-based talk show, adventure-based video game, tweet, or horror types, that John Brown has a high interest level at Time 4. Conversely, interest module 108 may determine that for the same sensor data at Time 4 for a drama, melodrama, or classical music, that John Brown has a low interest level at Time 4. This can be performed based on the sensor data without first determining an engagement level or state, though this may also be performed.

Block 408, either after block 404 or 406, provides the demographics, identity, engagement, state, and/or interest level. State module 106 or interest module 108 may provide this information to various entities. These entities can be any of the above-mentioned advertisers, media creators, and media providers. Providing this information to an advertising entity or media provider can be effective to enable the advertising entity to measure a value of their advertisements shown during a media program or the media provider to set advertisement costs. Providing this information to a media creator can be effective to enable the media creator to assess a potential value of a similar media program or portion thereof. For example, a media creator, prior to releasing the media program to the general public, may determine portions of the media program that are not well received, and thus alter the media program to improve it.

Further, this information may be provided to other entities as well. Providing this information to a rating entity, for example, can be effective to enable the rating entity to automatically rate the media program for the person (e.g., four stars out of five or a “thumbs up”). Providing this information to a media controller, for example, may enable the media controller to improve media control and presentation, such as by pausing the media program responsive to all of the persons in the audience departing the room.

As noted herein, the techniques can determine numerous states for a person over the course of most media programs, even for 15-second advertisements or video snippets. In such a case block 404 is repeated, such as at one-second periods.

Furthermore, state module 106 may determine not only multiple states for a person over time, but also various different states at a particular time. A person may be both laughing and looking away, for example, both of which are states that may be determined and provided or used to determine the persons' interest level.

Further still, either or both of state module 106 and interest module 108 may determine engagement, states, and/or interest levels based on historical data in addition to sensor data or media type. In one case a person's historical sensor data is used to normalize the person's engagement, states, or interest levels. If, for example, Susan Brown is viewing a media program and sensor data for her is received, the techniques may normalize or otherwise learn how best to determine engagement, states, and interest levels for her based on her historical sensor data. If Susan Brown's historical sensor data indicates that she is not a particularly expressive or vocal person, the techniques may adjust for this history. Thus, lower-amplitude audio may be sufficient to determine that Susan Brown laughed compared to an amplitude of audio used to determine that a typical person laughed.

In another case historical engagement, states, or interest levels of the person for which sensor data is received are compared with historical engagement, states, or interest levels for other people. Thus, a lower interest level may be determined for Lydia Brown based on data indicating that she exhibits a high interest for almost every media program she watches compared to other people's interest levels (either generally or for the same media program). In either of these cases the techniques learn over time, and thereby can normalize engagement, states, and/or interest levels.

The preceding discussion describes methods relating to determining an audience's state or interest using passive sensor data. Aspects of these methods may be implemented in hardware (e.g., fixed logic circuitry), firmware, software, manual processing, or any combination thereof. A software implementation represents program code that performs specified tasks when executed by a computer processor. The example methods may be described in the general context of computer-executable instructions, which can include software, applications, routines, programs, objects, components, data structures, procedures, modules, functions, and the like. The program code can be stored in one or more computer-readable memory devices, both local and/or remote to a computer processor. The methods may also be practiced in a distributed computing mode by multiple computing devices. Further, the features described herein are platform-independent and can be implemented on a variety of computing platforms having a variety of processors.

These techniques may be embodied on one or more of the entities shown in FIGS. 1-3 and 6 (device 600 is described below), which may be further divided, combined, and so on. Thus, these figures illustrate some of many possible systems or apparatuses capable of employing the described techniques. The entities of these figures generally represent software, firmware, hardware, whole devices or networks, or a combination thereof. In the case of a software implementation, for instance, the entities (e.g., state module 106 and interest module 108) represent program code that performs specified tasks when executed on a processor (e.g., processor(s) 204 and/or 306). The program code can be stored in one or more computer-readable memory devices, such as media 206 and/or 308 or computer-readable media 614 of FIG. 6.

Example Device

FIG. 6 illustrates various components of example device 600 that can be implemented as any type of client, server, and/or computing device as described with reference to the previous FIGS. 1-5 to implement techniques for determining audience state or interest using passive sensor data. In embodiments, device 600 can be implemented as one or a combination of a wired and/or wireless device, as a form of television mobile computing device (e.g., television set-top box, digital video recorder (DVR), etc.), consumer device, computer device, server device, portable computer device, user device, communication device, video processing and/or rendering device, appliance device, gaming device, electronic device, System-on-Chip (SoC), and/or as another type of device. Device 600 may also be associated with a user (e.g., a person) and/or an entity that operates the device such that a device describes logical devices that include users, software, firmware, and/or a combination of devices.

Device 600 includes communication devices 602 that enable wired and/or wireless communication of device data 604 (e.g., received data, data that is being received, data scheduled for broadcast, data packets of the data, etc.). The device data 604 or other device content can include configuration settings of the device, media content stored on the device (e.g., media programs 210), and/or information associated with a user of the device. Media content stored on device 600 can include any type of audio, video, and/or image data. Device 600 includes one or more data inputs 606 via which any type of data, media content, and/or inputs can be received, such as human utterances, user-selectable inputs, messages, music, television media content, recorded video content, and any other type of audio, video, and/or image data received from any content and/or data source.

Device 600 also includes communication interfaces 608, which can be implemented as any one or more of a serial and/or parallel interface, a wireless interface, any type of network interface, a modem, and as any other type of communication interface. The communication interfaces 608 provide a connection and/or communication links between device 600 and a communication network by which other electronic, computing, and communication devices communicate data with device 600.

Device 600 includes one or more processors 610 (e.g., any of microprocessors, controllers, and the like), which process various computer-executable instructions to control the operation of device 600 and to enable techniques for determining audience state or interest using passive sensor data. Alternatively or in addition, device 600 can be implemented with any one or combination of hardware, firmware, or fixed logic circuitry that is implemented in connection with processing and control circuits, which are generally identified at 612. Although not shown, device 600 can include a system bus or data transfer system that couples the various components within the device. A system bus can include any one or combination of different bus structures, such as a memory bus or memory controller, a peripheral bus, a universal serial bus, and/or a processor or local bus that utilizes any of a variety of bus architectures.

Device 600 also includes computer-readable storage media 614, such as one or more memory devices that enable persistent and/or non-transitory data storage (i.e., in contrast to mere signal transmission), examples of which include random access memory (RAM), non-volatile memory (e.g., any one or more of a read-only memory (ROM), flash memory, EPROM, EEPROM, etc.), and a disk storage device. A disk storage device may be implemented as any type of magnetic or optical storage device, such as a hard disk drive, a recordable and/or rewriteable compact disc (CD), any type of a digital versatile disc (DVD), and the like. Device 600 can also include a mass storage media device 616.

Computer-readable storage media 614 provides data storage mechanisms to store the device data 604, as well as various device applications 618 and any other types of information and/or data related to operational aspects of device 600. For example, an operating system 620 can be maintained as a computer application with the computer-readable storage media 614 and executed on processors 610. The device applications 618 may include a device manager, such as any form of a control application, software application, signal-processing and control module, code that is native to a particular device, a hardware abstraction layer for a particular device, and so on.

The device applications 618 also include any system components, engines, or modules to implement techniques for determining audience state or interest using passive sensor data. In this example, the device applications 618 can include state module 106 and interest module 108.

CONCLUSION

Although embodiments of techniques and apparatuses for determining an audience's state or interest using passive sensor data have been described in language specific to features and/or methods, it is to be understood that the subject of the appended claims is not necessarily limited to the specific features or methods described. Rather, the specific features and methods are disclosed as example implementations for determining an audience's state or interest using passive sensor data.

Claims

1. A computer-implemented method comprising: causing a microphone to capture sensed audio originating from an audience during presentation of a media program to the audience;accessing previously-recorded audio data for an individual member of the audience;comparing an individual amplitude of the previously-recorded audio data for the individual member of the audience to other amplitudes of other users;in an instance when the individual amplitude of the individual member of the audience has a lower magnitude than the other amplitudes of the other users, determining that the individual member of the audience is relatively less expressive or vocal than a typical user;determining an adjusted interest level of the individual member of the audience based at least on the sensed audio, the adjusted interest level accounting for the individual member of the audience being relatively less expressive or vocal than a typical user; andoutputting a representation of the adjusted interest level of the individual member of the audience.

2. The computer-implemented method of claim 1, further comprising: determining, based at least on the sensed audio, orientation sufficient to determine a portion of the sensed audio originating from the individual member of the audience.

3. The computer-implemented method of claim 1, further comprising: in an instance when the individual amplitude of the individual member of the audience is insufficient to map to a particular interest level for a typical user, determining that the individual member of the audience has the particular interest level based at least on the individual member of the audience being relatively less expressive or vocal than a typical user.

4. The computer-implemented method of claim 1, further comprising: using a depth sensor, obtaining depth data for the audience;processing the depth data to determine a relative deviation between a head orientation of the individual member of the audience and a presentation device presenting the media program; anddetermining the adjusted interest level based at least on relative deviation.

5. The computer-implemented method of claim 1, further comprising: sensing a skeletal orientation of an upper torso of the individual member of the audience relative to a lower torso of the individual member of the audience;based at least on the skeletal orientation, detecting that the individual member of the audience is leaning toward a presentation device presenting the media program; anddetermining the adjusted interest level based at least on the detecting that the individual member of the audience is leaning toward the presentation device.

6. The computer-implemented method of claim 1, where the adjusted interest level of the individual member of the audience is associated with a particular portion of the media program having a corresponding media category.

7. A method comprising: associating a laughing user state with different interest levels for different media genres, including a first interest level for a comedy genre and a second interest level for a drama genre, the first interest level being relatively higher than the second interest level;receiving sensed audio from a microphone, the sensed audio originating from an audience during presentation of media to the audience;receiving other sensor data from one or more other sensors that sense the audience during presentation of the media;determining respective media genres of different time periods of the media, including a first time period associated with the comedy genre and a second time period associated with the drama genre;based at least on the sensed audio received from the microphone and the other sensor data, determining that an individual member of the audience exhibits the laughing user state during the first time period and the second time period;determining that the individual member of the audience has the first interest level when the individual member exhibits the laughing user state during the first time period associated with the comedy genre and that the individual member has the second interest level when the individual member exhibits the laughing user state during the second time period associated with the drama genre; andoutputting the first interest level in association with the first time period and the second interest level in association with the second time period.

8. The method of claim 7, wherein the media comprises a movie, the method further comprising: determining that a first portion of the movie that is presented during the first time period is designated as the comedy genre; anddetermining that a second portion of the movie that is presented during the second time period is designated as the drama genre.

9. The method of claim 7, further comprising: associating a screaming user state with a high interest level for a horror genre; anddetermining that the individual member of the audience has the high interest level when the individual member exhibits the screaming user state during a particular time period associated with the horror genre.

10. The method of claim 7, performed by a computing device located remotely from the audience.

11. The method of claim 7, further comprising: during a third time period, detecting that the individual member of the audience exhibits a cheering user state when skeletal movement of the individual member includes arm movement and the sensed audio has a high amplitude.

12. The method of claim 11, further comprising: distinguishing between the cheering user state and a clapping user state based at least on a type of arm movement by the individual member of the audience.

13. The method of claim 7, further comprising: detecting that the individual member of the audience exhibits the laughing user state during the first time period based at least on head movement and facial feature changes of the individual member during the first time period.

14. A system comprising: one or more processors; andone or more memories storing computer-executable instructions which, when executed by the one or more processors, cause the one or more processors to:obtain sensed audio originating from an audience during presentation of a media program to the audience;access previously-recorded audio data for an individual member of the audience;compare an individual amplitude of the previously-recorded audio data for the individual member of the audience to other amplitudes of other users;in an instance when the individual amplitude of the individual member of the audience has a lower magnitude than the other amplitudes of the other users, determine that the individual member of the audience is relatively less expressive or vocal than a typical user;determine an adjusted interest level of the individual member of the audience based at least on the sensed audio, the adjusted interest level accounting for the individual member of the audience being relatively less expressive or vocal than a typical user; andoutput a representation of the adjusted interest level of the individual member of the audience.

15. The system of claim 14, wherein the computer-executable instructions, when executed by the one or more processors, cause the one or more processors to: determine, based at least on the sensed audio, orientation sufficient to identify a portion of the sensed audio that originates from the individual member of the audience.

16. The system of claim 14, wherein the computer-executable instructions, when executed by the one or more processors, cause the one or more processors to: in an instance when the individual amplitude of the individual member of the audience is insufficient to map to a particular interest level for a typical user, determine that the individual member of the audience has the particular interest level based at least on the individual member of the audience being relatively less expressive or vocal than a typical user.

17. The system of claim 14, wherein the computer-executable instructions, when executed by the one or more processors, cause the one or more processors to: obtain depth data for the audience;process the depth data to determine a relative deviation between a head orientation of the individual member of the audience and a presentation device presenting the media program; anddetermine the adjusted interest level based at least on relative deviation.

18. The system of claim 14, wherein the adjusted interest level of the individual member of the audience is associated with a particular portion of the media program having a corresponding media category.

19. The system of claim 14, further comprising a microphone configured to sense the audio originating from the audience.

20. The system of claim 14, provided in a remote computing device having access to the sensed audio over a communication network.