Claims:We claim:

1. A media content recommendation system (100), comprising:
a media player (204) configured to capture an interactive action performed by a user (112A) during playback of a particular portion of a selected media content being played on a media device (102A), wherein the selected media content comprises different portions, each of the different portions belonging to a particular category, wherein the interactive action comprises one of a positive interactive action and a negative interactive action;
an audio sensor (224) configured to detect and record an ambient audio during playback of the selected media content;
an event identification system (222) that is operatively coupled to the audio sensor (224), wherein the event identification system (222) analyzes the ambient audio and identifies occurrence of an event within a particular time period prior to the negative interactive action performed by the user (112A) upon identifying that the ambient audio comprises previously stored audio unrelated to the selected media content; and
a recommendation subsystem (110) that is communicatively coupled to the media player (204) and the event identification system (222), wherein the recommendation subsystem (110) is configured to:
assign a first weightage to the negative interactive action only when the event identification system (222) identifies occurrence of the event within the particular time period prior to the negative interactive action performed by the user (112A), and assign a second weightage different from the first weightage to the negative interactive action when the event identification system (222) fails to identify occurrence of the event within the particular time period prior to the negative interactive action performed by the user (112A);
determine a category rating for the particular category associated with the different portions of the selected media content based on at least the first weightage or the second weightage;
identify a list of related media content belonging to the particular category of the selected media content from a content server (106) when the determined category rating is above a particular threshold; and
automatically transmit the identified list of related media content to the media device (102A) via a communications link (116) as personalized content recommendation when a subsequent viewing session is initiated by the same user (112A).

2. The media content recommendation system (100) as claimed in claim 1, further comprising:
an imaging sensor (208) that captures one or more images of a viewing area in which the media device (102A) is placed, wherein the imaging sensor (208) comprises a camera, and wherein the audio sensor comprises a microphone, and
an image processing system (218) that is adapted to:
identify the user (112A) currently watching the selected media content based on facial information obtained from the captured images and facial information of the user (112A) stored in a user profile storage system (212) of the media device (102A); and
identify a gaze duration that indicates an amount of time the user (112A) gazed at the media device (102A) when the media device (102A) plays the particular portion of the selected media content by analyzing the captured images, wherein the recommendation subsystem (110) determines the engagement score based on the identified gaze duration, a total duration of the particular portion, and a duration of the particular portion watched by the user (112A).

3. The media content recommendation system (100) as claimed in claim 2, further comprising an emotion identification system (222) that is configured to identify emotions of the user (112A) while watching the different portions of the selected media content based on measurements captured by one or more sensors, wherein the one or more sensors comprise a motion sensor (206), the imaging sensor (208), a temperature sensor, an infrared sensor, a pulse rate sensor, an electrocardiogram sensor, and an electroencephalography sensor, wherein the motion sensor (206) comprises one or more of a thermo-graphic camera, a depth-sensing camera, an ultrasonic sensor, and a light detection and ranging sensor.

4. The media content recommendation system (100) as claimed in claim 1, wherein the recommendation subsystem (110) resides in a content server (106), wherein the content server (106) comprises a video-on-demand (VOD) server, an over-the-top (OTT) server, or a server associated with a cable provider, a satellite provider, or a broadcast provider.

5. The media content recommendation system (100) as claimed in claim 1, wherein the media content recommendation system (100) is implemented in one or more of an entertainment content recommendation system, an advertisement recommendation system, an infotainment system of a vehicle, and an online classroom system.

6. A method for recommending media content, comprising:
capturing an interactive action performed by a user (112A) during playback of a particular portion of a selected media content being played by a media player (204) on a media device (102A), wherein the selected media content comprises different portions, each of the different portions belonging to a particular category, wherein the interactive action comprises one of a positive interactive action and a negative interactive action;
detecting and recording an ambient audio during playback of the selected media content using an audio sensor (224);
analyzing the ambient audio and identifying occurrence of an event within a particular time period prior to the negative interactive action performed by the user (112A) by an event identification system (222) upon identifying that the ambient audio comprises a previously stored audio unrelated to the selected media content;
assigning a first weightage to the negative interactive action only when the event identification system (222) identifies occurrence of the event within the particular time period prior to the negative interactive action performed by the user (112A), and assigning a second weightage different from the first weightage to the negative interactive action when the event identification system (222) fails to identify occurrence of the event within the particular time period prior to the negative interactive action performed by the user (112A);
determining a category rating for the particular category associated with the different portions of the selected media content by the recommendation subsystem (110) based on at least the first weightage or the second weightage;
identifying a list of related media content belonging to the particular category of the selected media content from a content server (106) when the determined category rating is above a particular threshold; and
automatically transmitting the identified list of media content to the media device (102A) via a communications link (116) as personalized content recommendation when a subsequent viewing session is initiated by the same user (112A).

7. The method as claimed in claim 6, further comprising:
capturing one or more images of a viewing area in which the media device (102A) is placed by an imaging sensor (208);
identify the user (112A) currently watching the selected media content based on facial information obtained from the captured images and facial information of the user (112A) stored in a user profile storage system (212) of the media device (102A);
identifying an emotion of the user (112A) during playback of each portion of the selected media content by processing the captured images; and
capturing an interactive action performed by the user (112A) during playback of each portion of the selected media content,
wherein the positive interactive action comprises one of resuming playback of the selected media content, replaying one or more portions of the selected media content, and seeking a specific portion of the selected media content, and wherein the negative interactive action comprises one of pausing playback of the selected media content, forwarding one or more portions of the selected media content, terminating playback of the selected media content, and navigating from the selected media content to another media content before the media player (204) completes playback of the selected media content.

8. The method as claimed in claim 7, further comprising:
identifying a gaze duration for each portion of the selected media content by analyzing the captured images using an image processing system (218) coupled to the imaging sensor (208), wherein the identified gaze duration associated with the particular portion of the selected media content indicates an amount of time the user (112A) gazed at the media device (102A) when the media device (102A) plays the particular portion; and
determining an engagement score that indicates an engagement level of the user (112A) with each of the different portions of the selected media content, wherein the recommendation subsystem (110) determines the engagement score for the particular portion of the selected media content based on the identified gaze duration, a total duration of the particular portion, and a duration of the particular portion watched by the user (112A).

9. The method as claimed in claim 8, further comprising:
processing the ambient audio detected by the audio sensor (224) using the event identification system (222) to identify a set of acoustic features associated with the ambient audio, wherein the identified acoustic features comprises one or more of mel-frequency, cepstral coefficients, spectral energy, spectral entropy, spectral flatness, and spectral flux;
comparing the identified acoustic features with the previously stored audio unrelated to the selected media content that is stored in the user profile storage system (212), wherein the previously stored audio comprises one or more of a doorbell sound, a phone ringtone, a whistle produced by a pressure cooker, an oven timer alert, and a human voice;
determining if the identified acoustic features match with acoustic features associated with the previously stored audio; and
identifying the event that occurred while watching the particular portion of the selected media content based on the determined acoustic features that match acoustic features associated with previously stored audio unrelated to the selected media content.

10. The method as claimed in claim 9, further comprising:
determining an initial weightage for a category associated with a portion of the selected media content by the recommendation subsystem (110) based on a default value stored for the category in the weightage database (118);
determining an average weightage for each of the different portions based on the associated duration, the associated initial weightage, and a sum of initial weightages determined for the different portions of the selected media content; and
determining an actual weightage for each of the different portions based on the determined average weightage and a sum of average weightages determined for different portions of the selected media content.

11. The method as claimed in claim 10, further comprising:
determining a watch weightage for a portion that comprises opening credits or closing credits of the selected media content based on a default value stored in a weightage database (118) of the recommendation subsystem (110); and
determining the watch weightage for a portion of the selected media content that lacks opening credits or closing credits based on another default value stored in the weightage database (118), an interactive action performed by the user (112A) while watching the portion, identification of an occurrence of the event within the particular time period prior to the interactive action performed by the user (112A), and a weightage of the interactive action stored in the weightage database (118).

12. The method as claimed in claim 11, further comprising:
classifying the particular category associated with the selected media content into one of a highly preferred category when the associated category rating is above a first threshold, a moderately preferred category when the category rating is in between the first threshold and a second threshold, and a less preferred category when the category rating is below the second threshold; and
generating a preference profile for the user (112A) by the recommendation subsystem (110), wherein the generated preference profile comprises the particular category classified under one of the highly preferred category, moderately preferred category, and less preferred category.

13. The method as claimed in claim 12, further comprising generating preference profiles for a plurality of other users (112B-N) of media devices (102A-N), wherein the generated preference profiles comprise categories of media content preferred by the users (112B-N) identified based on a set of information received from their corresponding media devices (102A-N), wherein the set of information received from the corresponding media devices (102A-N) comprises emotions of the users (112B-N) while watching a plurality of media content, interactive actions performed by the users (112B-N) while watching the plurality of media content, occurrence of events while watching the plurality of media content, and engagement levels of the users (112B-N) with the plurality of media content.

14. The method as claimed in claim 13, further comprising:
identifying a set of users (112A-C) watching media content presented by the media device (102A) together as a group by the imaging sensor (208);
identifying a category of media content that is preferred by all users (112A-C) in the group by the recommendation subsystem (110) based on preference profiles of the users (112A-C) stored in a preference database (114); and
transmitting a set of media content belonging to the identified category from a content server (106) to the media device (102A) via the communications link (116) as personalized content recommendation for the group.

, Description:CONTENT RECOMMENDATION SYSTEM FOR A MEDIA DEVICE

RELATED ART

[0001] Embodiments of the present disclosure relate generally to content recommendation. More particularly, the present disclosure relates to a system and method for recommending media content to a group of users watching the media content together.
[0002] Presently, a vast amount of media content is available for user consumption via various platforms such as via a cable television, satellite television, a broadcast television, a video on demand service, an over-the-top service, and/or the internet. The media content available on these platforms, for example, includes television shows, programs, movies, clips, music, live content, and sport highlights. As the media content available on these platforms continues to proliferate, it is increasingly difficult for users to select and watch appropriate media content that suit their preferences.
[0003] Accordingly, these platforms employ recommendation systems that predict and recommend media content that users may want to watch. Specifically, certain presently existing recommendation systems recommend media content to a user based on a watch history of the user. For example, US patent US9536246B2 describes a content recommendation system that transmits a list of recommended content to a user based on the watch history of the user. Specifically, this content recommendation system sets pausing of the media content by the user as a negative evaluation input, and watching the media content to the end as a positive evaluation input. The content recommendation system subsequently provides recommendations to the user for media content that is similar to the media content corresponding to the positive evaluation input and does not recommend media content similar to the media content corresponding to the negative evaluation input. However, these user actions may be inadvertent or irrelevant, such as pausing the media content on receiving a phone call, and may not accurately indicate actual user preferences. Additionally, such existing recommendation systems cannot effectively integrate multiple users’ preferences and recommend media content that satisfies preferences of all users in the group.
[0004] Accordingly, certain content recommendation systems attempt to recommend content suitable for group viewing when a user watches media content with other users. For example, certain existing recommendation systems allow a user such as a parent to set a group preference for watching media content together with family members. However, the recommended media content, identified based on the group preference set previously, may not satisfy preferences of all family members, for example grandparents and kids, in the family. Additionally, preferences of different individuals in the family may also dynamically change during different times, for example, during holidays, weekends, or when feeling unwell.
[0005] Accordingly, there is a need for an improved recommendation system that accurately recommends media content to a single user or a group of users watching media content together, and satisfies preferences of all users in the group.

BRIEF DESCRIPTION

[0006] It is an objective of the present disclosure to provide a media content recommendation system. The media content recommendation system includes a media player, an audio sensor, an event identification system, and a recommendation subsystem. The media player configured to capture an interactive action performed by a user during playback of a particular portion of a selected media content being played on a media device. The selected media content includes different portions. Each of the different portions belonging to a particular category. The interactive action includes one of a positive interactive action and a negative interactive action. The audio sensor configured to detect and record an ambient audio during playback of the selected media content. The event identification system is operatively coupled to the audio sensor. The event identification system analyzes the ambient audio and identifies occurrence of an event within a particular time period prior to the negative interactive action performed by the user upon identifying that the ambient audio includes previously stored audio unrelated to the selected media content.
[0007] The recommendation subsystem is communicatively coupled to the media player and the event identification system. The recommendation subsystem assigns a first weightage to the negative interactive action only when the event identification system identifies occurrence of the event within the particular time period prior to the negative interactive action performed by the user. The recommendation subsystem assigns a second weightage different from the first weightage to the negative interactive action when the event identification system fails to identify occurrence of the event within the particular time period prior to the negative interactive action performed by the user. In addition, the recommendation subsystem determines a category rating for the particular category associated with the different portions of the selected media content based on at least the first weightage or the second weightage.
[0008] Furthermore, the recommendation subsystem identifies a list of related media content belonging to the particular category of the selected media content from a content server when the determined category rating is above a particular threshold. Additionally, the recommendation subsystem automatically transmits the identified list of related media content to the media device via a communications link as personalized content recommendation when a subsequent viewing session is initiated by the same user. The media content recommendation system further includes an imaging sensor. The imaging sensor captures one or more images of a viewing area in which the media device is placed. The imaging sensor includes a camera. The audio sensor includes a microphone. An image processing system identifies the user currently watching the selected media content based on facial information obtained from the captured images and facial information of the user stored in a user information storage system of the media device. The image processing system further identifies a gaze duration that indicates an amount of time the user gazed at the media device when the media device plays the particular portion of the selected media content by analyzing the captured images. The recommendation subsystem determines the engagement score based on the identified gaze duration, a total duration of the particular portion, and a duration of the particular portion watched by the user.
[0009] An emotion identification system identifies emotions of the user while watching the different portions of the selected media content based on measurements captured by one or more sensors. The one or more sensors include a motion sensor, the imaging sensor, a temperature sensor, an infrared sensor, a pulse rate sensor, an electrocardiogram sensor, and an electroencephalography sensor. The motion sensor includes one or more of a thermo-graphic camera, a depth-sensing camera, an ultrasonic sensor, and a light detection and ranging sensor. The recommendation subsystem resides in a content server. The content server includes a video-on-demand (VOD) server, an over-the-top (OTT) server, or a server associated with a cable provider, a satellite provider, or a broadcast provider. The media content recommendation system is implemented in one or more of an entertainment content recommendation system, an advertisement recommendation system, an infotainment system of a vehicle, and an online classroom system.
[0010] It is another objective of the present disclosure to provide a method for recommending media content. The method includes capturing an interactive action performed by a user during playback of a particular portion of a selected media content being played by a media player on a media device. The selected media content includes different portions. Each of the different portions belonging to a particular category. The interactive action includes one of a positive interactive action and a negative interactive action. Further, the method includes detecting and recording an ambient audio during playback of the selected media content using an audio sensor. Furthermore, the method includes analyzing the ambient audio and identifying occurrence of an event within a particular time period prior to the negative interactive action performed by the user by an event identification system upon identifying that the ambient audio comprises a previously stored audio unrelated to the selected media content. In addition, the method includes assigning a first weightage to the negative interactive action only when an event identification system identifies occurrence of the event within the particular time prior to the negative interactive action performed by the user.
[0011] The method further includes assigning a second weightage different from the first weightage to the negative interactive action when the event identification system fails to identify occurrence of the event within the particular time period prior to the negative interactive action performed by the user. Further, the method includes determining a category rating for the particular category associated with the different portions of the selected media content by a recommendation subsystem based on at least the first weightage or the second weightage. Furthermore, the method includes identifying a list of related media content belonging to the particular category of the selected media content from a content server when the determined category rating is above a particular threshold. The method further includes automatically transmitting the identified list of media content to the media device via a communications link as personalized content recommendation when a subsequent viewing session is initiated by the same user.
[0012] The method includes capturing one or more images of a viewing area in which the media device is placed by an imaging sensor. The user currently watching the selected media content is identified based on facial information obtained from the captured images and facial information of the user stored in a user information storage system of the media device. An emotion of the user during playback of each portion of the selected media content is identified by processing the captured images. An interactive action performed by the user during playback of each portion of the selected media content is captured. The positive interactive action includes one of resuming playback of the selected media content, replaying one or more portions of the selected media content, and seeking a specific portion of the selected media content. The negative interactive action includes one of pausing playback of the selected media content, forwarding one or more portions of the selected media content, terminating playback of the selected media content, and navigating from the selected media content to another media content before the media player completes playback of the selected media content.
[0013] The method includes identifying a gaze duration for each portion of the selected media content by analyzing the captured images using an image processing system coupled to the imaging sensor. The identified gaze duration associated with the particular portion of the selected media content indicates an amount of time the user gazed at the media device when the media device plays the particular portion. The method further includes determining an engagement score that indicates an engagement level of the user with each of the different portions of the selected media content. The recommendation subsystem determines the engagement score for the particular portion of the selected media content based on the identified gaze duration, a total duration of the particular portion, and a duration of the particular portion watched by the user. Further, the method includes processing the ambient audio detected by the audio sensor using the event identification system to identify a set of acoustic features associated with the ambient audio. The identified acoustic features includes one or more of mel-frequency, cepstral coefficients, spectral energy, spectral entropy, spectral flatness, and spectral flux.
[0014] Furthermore, the method includes comparing the identified acoustic features with the previously stored audio unrelated to the selected media content that is stored in the user profile storage system. The previously stored audio includes one or more of a doorbell sound, a phone ringtone, a whistle produced by a pressure cooker, an oven timer alert, and a human voice. Furthermore, the method includes determining if the identified acoustic features match with acoustic features associated with the previously stored audio. In addition, the method includes identifying the event that occurred while watching the particular portion of the selected media content based on the determined acoustic features that match acoustic features associated with previously stored audio unrelated to the selected media content.
[0015] Further, the method includes determining an initial weightage for a category associated with a portion of the selected media content by the recommendation subsystem based on a default value stored for the category in the weightage database. Furthermore, the method includes determining an average weightage for each of the different portions based on the associated duration, the associated initial weightage, and a sum of initial weightages determined for the different portions of the selected media content. In addition, the method includes determining an actual weightage for each of the different portions based on the determined average weightage and a sum of average weightages determined for different portions of the selected media content. The method further includes determining a watch weightage for a portion that includes opening credits or closing credits of the selected media content based on a default value stored in a default value storage system of the recommendation subsystem. The method includes determining the watch weightage for a portion of the selected media content that lacks opening credits or closing credits. The recommendation subsystem determines the watch weightage based on another default value stored in the weightage database, an interactive action performed by the user while watching the portion, identification of an occurrence of the event within the particular time period prior to the interactive action, and a weightage of the interactive action stored in the weightage database.
[0016] The method includes classifying the particular category into one of a highly preferred category when the associated category rating is above a first threshold, a moderately preferred category when the category rating is in between the first threshold and a second threshold, and a less preferred category when the category rating is below the second threshold. The method further includes generating a preference profile for the user by the recommendation subsystem. The generated preference profile includes the particular category classified under one of the highly preferred category, moderately preferred category, and less preferred category. In addition, the method includes generating preference profiles for a plurality of other users of media devices. The generated preference profiles include categories of media content preferred by the users identified based on a set of information received from their corresponding media devices. The set of information received from the corresponding media devices includes emotions of the users while watching a plurality of media content, interactive actions performed by the users while watching the media content, occurrence of events while watching the media content, and engagement levels of the users with the media content. In addition, the method includes identifying a set of users watching media content presented by the media device together as a group by an imaging sensor. Further, the method includes identifying a category of media content that is preferred by all users in the group by the recommendation subsystem based on preference profiles of the users stored in a preference database. Furthermore, the method includes transmitting a set of media content belonging to the identified category from a content server to the media device via the communications link as personalized content recommendation for the group.

BRIEF DESCRIPTION OF DRAWINGS

[0017] These and other features, aspects, and advantages of the claimed subject matter will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
[0018] FIG. 1 illustrates a block diagram depicting an exemplary media content recommendation system that transmits recommendations including one or more media content to a set of media devices deployed at different customer premises, in accordance with aspects of the present disclosure;
[0019] FIG. 2 illustrates a block diagram depicting an exemplary media device associated with a user that receives recommendations provided by the media content recommendation system of FIG. 1, in accordance with aspects of the present disclosure;
[0020] FIGS. 3A-C illustrate a flow diagram depicting an exemplary method for generating a preference profile that indicates media content preferences of a user using the media content recommendation system of FIG. 1, in accordance with aspects of the present disclosure; and
[0021] FIGS. 4A-B illustrate a flow diagram depicting an exemplary method for providing content recommendation to one or more users watching a video together as a group using the media content recommendation system of FIG. 1, in accordance with aspects of the present disclosure.

DETAILED DESCRIPTION

[0022] The following description presents an exemplary system and associated method for providing content recommendation. Particularly, embodiments described herein disclose a media content recommendation system that recommends media content to a group of users, for example, family members watching media content together. To that end, the recommendation system initially identifies media content preferences of each user in a family and stores the identified media content preferences in a preference database. For example, the recommendation system uses one or more sensors that identify emotions and various interactive actions performed by a user in the family while watching a selected video to identify users’ media content preferences.
[0023] Further, the recommendation system categorizes the interactive actions performed by the user into one of a positive input and a negative input. For example, the recommendation system categorizes the interactive actions such as watching a portion of the selected video completely and replaying the watched portion as a positive input. In addition, the recommendation system categorizes the interactive actions such as forwarding the selected video, pausing the selected video, skipping one or more portions of the selected video, and navigating from the selected video to another video without completely watching the selected video as a negative input.
[0024] In certain embodiments, the recommendation system also identifies the context behind negative inputs provided by the user. To that end, the recommendation system includes one or more ambient monitoring sensors, for example, an audio sensor that detects ambient audio that is generated during playback of the selected video. The recommendation system then analyses the ambient audio to identify if the ambient audio includes both audio elements that are related to the selected video and unrelated audio elements that match a previously stored audio relating to an unrelated event such as a doorbell, an oven times, a phone ringer, or voice of a family member or known friend. The recommendation system then correlates the previously stored audio detected from the ambient audio with an input provided by the user to identify the context behind the negative input provided by the user. For example, the audio sensor detects ambient audio that includes both audio related to the selected video and a previously stored doorbell sound when the user is watching the selected video. The recommendation system subsequently identifies a pause action performed by the user. In this example, the recommendation system identifies that the context behind pausing playback of the selected video is to attend to the person waiting outside the home. Further, in this example, the recommendation system identifies that the user is interested in the selected video when the user replays the video, even though he or she previously paused playback of the selected video, and provides future recommendations including videos that are similar to the selected video.
[0025] However, in similar scenarios, conventional recommendation systems do not identify the context behind the negative input of pausing playback of the selected video. Instead, the conventional recommendation systems identify the pausing action as a negative input and determine that the user is not interested in the selected video, and thus, may not provide future recommendations that are similar to the selected video. Accordingly, when compared to the conventional recommendation systems, the present recommendation system provides more accurate recommendations to both a single user and a group of users watching the selected video together by identifying the context of user actions in the real world by monitoring additional events.
[0026] In addition to identifying context, the recommendation system determines a rating for a category associated with the selected video watched by the user, an overall rating for the selected video, and individual ratings for different portions of the selected video, as described in detail with reference to FIGS. 3A-C. Further, the recommendation system classifies videos watched by the user into one or more preference categories, and generates a preference profile that indicates media content preferences of the user. Similarly, the recommendation system generates preference profiles for all other users in the family, and stores their preference profiles in a preference database. Subsequently, the recommendation system recommends a list of media content to users in the family watching media content together as a group based on the generated preference profiles such that the recommended content satisfies preferences all users in the group, as described in detail with reference to FIGS. 4A-B.
[0027] It may be noted that different embodiments of the present recommendation system may be used in many different application areas or systems. For example, in a broadcast or OTT network, the recommendation system may reside in a video on demand (VOD) server, an over the top (OTT) server, or a content distribution server owned by a cable services provider, a satellite services provider, or a broadcast services provider. In certain embodiments, the recommendation system identifies media content preferences of all individuals in a family based on their emotions and various interactive actions performed by those individuals while watching different media content. The recommendation system then provides future recommendations for entertaining media content based on the identified media content preferences to satisfy preferences of all individuals in the family.
[0028] In another example, the recommendation system may recommend targeted advertisements to one or more users based on identified emotions and various interactive actions performed by the users while watching a plurality of advertisements in the past. In yet another example, the recommendation system identifies that there is a toll ahead on the road using data captured by onboard sensors such as GPS sensors or cameras. Further, the recommendation system identifies that a user has paused music played by a vehicle’s infotainment system. In this example, the recommendation system identifies that the user has paused music to make a toll payment and not because of his or her disinterest in the music. Accordingly, the recommendation system may continue to provide future recommendations including music that are similar to the paused music.
[0029] When used in an online classroom application, the recommendation system identifies that someone has pressed a doorbell at a student’s premises based on audio data captured by an audio sensor deployed at the student’s premises. Subsequently, the recommendation system identifies that the student has exited a viewing area where a media device is placed. In this example, the recommendation system identifies that the user has exited the viewing area to attend a person waiting outside the student’s premises and not because of his or her disinterest in a lesson taught by a faculty. Further, the recommendation system automatically records a portion of the lesson missed by the student and later recommends the recorded portion to the student. When used in a theatre application, the recommendation system identifies ratings for movies. When used in selecting a movie for a prestigious award, the recommendation system identifies ratings for various movies based on juries’ emotions and interactive actions performed by the juries while watching the movies, and selects and recommends a particular movie that is rated high for the prestigious award.
[0030] Thus, the present recommendation system can be used in a variety of application areas where media content such as videos, audios, images, text, and/or other visual content need to be recommended to a single user or a group of users. However, for clarity, an embodiment of the present recommendation system will be described in greater detail with reference to FIG. 1 for recommendation of videos to a single user or a group of users watching media content together via an OTT application.
[0031] FIG. 1 illustrates a block diagram depicting an exemplary media content recommendation system (100) that transmits recommendations including one or more media content to a set of media devices (102A-N) deployed at different customer premises (104A-N). To that end, the media content recommendation system (100) includes a content server (106). In one embodiment, the content sever (106) corresponds to a server owned by one or more of a cable media content provider, a satellite media content provider, a broadcast content provider, a VOD content provider, and an OTT content provider. Particularly, the content server (106) includes a content database (108) that stores a plurality of media content to be transmitted as recommendations to the media devices (102A-N).
[0032] Conventional recommendation systems rely on metadata and actions of user belonging to the same demographic group to identify user preferences and provide relevant content recommendations to individual users. However, these conventional content recommendation systems fail to account for false positive or false negative inputs caused by unrelated events occurring in the real world that provide context for user action such as pausing a video on receiving a phone call or continuing play of video due to the user falling asleep or leaving the room. Additionally, these systems fail to identify a group of users and to provide content recommendation suited to the cumulative group. Conventional recommendation systems, thus, cause the users to mindlessly browse through and watch portions of irrelevant content, wasting time and network resources, while perpetuating user dissatisfaction.
[0033] Unlike such conventional content recommendation systems, the content server (106) includes a recommendation subsystem (110) that identifies media content preferences of different users (112A-C, 112D-E, and 112N) present in the customer premises (104A-N) and generates preference profiles for those users (112A-C, 112D-E, and 112N), as described in detail described in detail with reference to FIGS. 3A-C. Further, the recommendation subsystem (110) stores the generated preference profiles in a preference database (114).
[0034] Subsequently, the recommendation subsystem (110) transmits recommendations including a list of media content based on the generated preference profiles to each of the media devices (102A-N) via a communications link (116). For example, the recommendation subsystem (110) transmits recommendations including a list of media content to a media device (102A) based on preference profiles generated by the recommendation subsystem (110) for one or more users (112A-C) currently watching media content presented by the media device (102A).
[0035] In one embodiment, the recommendation subsystem (110) may be implemented by suitable code on a processor-based system, such as a general-purpose or a special-purpose computer. Accordingly, the recommendation subsystem (110), for example, include one or more general-purpose processors, specialized processors, graphical processing units, microprocessors, programming logic arrays, field programming gate arrays, integrated circuits, systems on chips, and/or other suitable computing devices. In certain embodiments, the communications link (116), for example, corresponds to a satellite-based communications system, a cable system, an over-the-top system, and the internet.
[0036] In certain embodiments, the recommendation subsystem (110) is communicatively coupled to the media devices (102A-N) via the communications link (116) to receive information related to user emotions and interactive actions performed by the users (112A-N) while watching the media content presented by their respective media devices (102A-N). The recommendation subsystem (110) then uses the received information to generate preference profiles for the users (112A-N). An embodiment of the recommendation subsystem (110) that generates a preference profile for a user (112A) based on information received from a media device (102A) is described in greater detail with reference to FIG. 2.
[0037] FIG. 2 illustrates a block diagram depicting an exemplary media device (102A) associated with the user (112A) that is adapted to receive recommendations from the content server (106) of FIG. 1. In one embodiment, the media device (102A) includes a display unit (202) for displaying one or more media content received from the content server (106) via the communications link (116). Examples of the media content include one or more of video content, audio content, image content, text, and/or other mixed-media content. Examples of the media device (102A) include a television, a smartphone, an infotainment system, a desktop, a laptop, a tablet, or any other device that receives and displays media content.
[0038] In one embodiment, the media content recommendation system (100) includes a custom media player (204) that may be installed and run on the media device (102A) as a native or downloadable application. In certain embodiments, the media player (204) presents a list of media content on the display unit (202) of the media device (102A) for user selection. Examples of the media content include one or more of television channel programs, over-the-top (OTT) content, video-on-demand (VoD) content, and any other content accessible via a satellite network, a cable network, a broadcast network, the internet, and/or a local storage device such as a hard drive or a universal serial bus (USB) device. For example, upon selecting a particular video from the list presented by the media player (204), the media device (102A) transmits a request to the content server (106) for streaming the selected video. In response, the content server (106) transmits the selected video to the media device (102A) via the communications link (116).
[0039] In certain embodiments, the media device (102A) receives the selected video as multiple media segments from the content server (106) and the media player (204) queues the received media segments for playing on the media device (102A). Further, the media player (204) processes the received video to identify one or more metadata markings embedded within the received video. Examples of the metadata markings embedded within the received video include length of the video, cast of actors, a publication date of the video, genre, and other categories associated with the video. For example, the media player (204) receives a selected video of 7-minutes length from the content server (106). A first portion corresponding to a first 0.5-minute of the selected video has metadata markings that correspond to “opening credits.” A second portion corresponding to 0.5 to 1.5th minute of the selected video has metadata markings that correspond to “adventure.” A third portion corresponding to 0.5 to 6.5th minute of the selected video has metadata markings that correspond to “comedy.” Further, a fourth portion corresponding to the last 0.5-minute of the selected video has metadata markings that correspond to “closing credits.” In the previously noted example, the media player (204) identifies that the first and fourth portions contain content related to opening and closing credits, the second portion contains content related to adventure category, and the third portion contains content related to comedy category.
[0040] As previously noted, conventional content recommendation systems provide recommendations on a per user basis and are often fail to identify presence of a group of users in real-time, and therefore, are unable to provide content recommendation suited to the cumulative group. In order to address these limitations of conventional content recommendation systems, in certain embodiments, the present media content recommendation system (100) includes one or more ambient sensors (203). These ambient sensors (203) are operatively coupled to the media device (102A), and identify one or more of the users (112A-N) currently present in a viewing area in the customer premises (104A-N) who may be watching the selected video played by the media player (204). Examples of the ambient sensors (203) include a motion sensor (206) and an imaging sensor (208). In one embodiment, the motion sensor (206) includes a thermo-graphic camera, a depth-sensing camera, an ultrasonic sensor, or a light detection and ranging sensor that captures physical attributes of a user detected in the viewing area such as size, stature, physique, gait information, and other physical information of the user. Upon identifying physical attributes of the detected user, the motion sensor (206) provides the identified physical attributes to a motion-sensor control system (210).
[0041] The motion-sensor control system (210) compares the identified physical attributes with physical attributes previously stored for the users (112A-C) registered with the media device (102A) to identify the detected user currently watching the selected video. For example, the motion-sensor control system (210) identifies a user currently watching the selected video as the user (112A) when the identified physical attributes have a significant match, for example over 90 percent match, with physical attributes of the user (112A) stored in a user profile storage system (212). To that end, the user profile storage system (212) includes a storage system such as a hard drive, a universal serial bus (USB) flash drive, a secure digital card, and a solid-state drive.
[0042] In certain embodiments, the motion-sensor control system (210) transmits a control signal to an imaging-sensor control system (214) to activate the imaging sensor (208) when the motion-sensor control system (210) fails to identify the user detected in the viewing area. In one embodiment, the imaging sensor (208) is an optical sensor such as a camera positioned to capture images of the viewing area in the vicinity of the media device (102A). In one embodiment, both the camera (208) and the imaging-sensor control system (214) are integrated within the media device (102A). However, in another embodiment, the camera (208) and the imaging-sensor control system (214) may be deployed as an independent unit that is physically separate, but is coupled to the media device (102A) via the communications link (116).
[0043] Upon activating the camera (208), the camera captures one or more images of the viewing area, and provides the captured images as input to an image processing system (218). The image processing system (218) identifies a user in the captured images as the user (112A), for example, when there is more than 80 percent match between the facial information of the user obtained from the captured images and the facial information of the user (112A) stored in the user profile storage system (212) associated with the media device (102A).
[0044] Conventional recommendation systems rely on user actions such as providing ratings and pausing or replaying the media content to identify user preferences and provide relevant content recommendations to individual users. However, most users such as elderly users and kids do not voluntarily rate the media content they watch, and therefore, content ratings usually reflect the preferences of only a small subset of technically aware users. In order to address the aforementioned issues, the present media content recommendation system (100), in certain embodiments, includes an emotion detection system (220) that includes one or more emotion detection sensors and is operatively coupled to the media device (102A). Examples of the emotion detection sensors include the camera (208), a microphone, a temperature sensor, an infrared sensor, a pulse rate sensor, an ECG sensor, an EEG sensor. Specifically, the emotion detection sensors capture biometric parameters of the user (112) while the user watches the selected video. The emotion detection system (220) then uses the captured biometric parameters to detect emotions of the user (112A) during playback of various portions of the selected video. Examples of the biometric parameters captured by the emotion detection sensors include speech, body temperature, heart rate, electroencephalography (EEG) signal, electrocardiogram (ECG) signal, blood pressure, facial expressions, and the like of the user. Examples of the emotions detected by the emotion detection system (220) using the captured biometric parameters include neutral, anger, sadness, surprise, fear, disgust, and happiness.
[0045] Additionally, in certain embodiments, the media player (204) captures one or more interactive actions performed by the user (112A) during playback of various portions of the selected video. Further, the media player (204) categorizes those interactive actions into one of a positive input and a negative input. For example, the media player (204) captures and categorizes interactive actions such as replaying a portion of the selected video and restarting playback of the selected video as positive input. In addition, the media player (204) captures and categorizes interactive actions such as forwarding the selected video, pausing the selected video, skipping one or more portions of the selected video, and navigating from the selected video to another video without completely watching the selected video as negative input. Further, the media player (204) stores the interactive actions categorized as positive and negative inputs in the user profile storage system (212) of the media device (102A).
[0046] As previously noted, conventional content recommendation systems fail to account for false positive or false negative inputs caused by unrelated events occurring in the real world that provide context for user action such as pausing a video on receiving a phone call or continuing play of video due to the user falling asleep or leaving the room. Accordingly, in certain embodiments, the present media content recommendation system (100) includes an event identification system (222), operatively coupled to the media device (102A), that identifies occurrence of one or more events while the user (112A) is watching the selected video.
[0047] In one embodiment, the event identification system (222), motion-sensor control system (210), emotion detection system (220), imaging-sensor control system (214), and image processing system (218) may be implemented by suitable code on a processor-based system, such as a general-purpose or a special-purpose computer. Accordingly, the event identification system (222), motion-sensor control system (210), emotion detection system (220), imaging-sensor control system (214), and image processing system (218), for example, include one or more general-purpose processors, specialized processors, graphical processing units, microprocessors, programming logic arrays, field programming gate arrays, integrated circuits, system on chips, and/or other suitable computing devices.
[0048] In one embodiment, the event identification system (222) is operatively coupled to one or more of the motion sensor (206), the imaging sensor (208), and an audio sensor (224) to identify events occurring in real-time while the user is watching the selected video content. By way of example, the audio sensor (224) such as a microphone, captures ambient audio detected during playback of the selected video. In this example, the event identification system (222) analyzes the detected ambient audio to identify if the ambient audio includes both audio that is related to the selected video and a previously stored audio that is unrelated to the selected video. To that end, the event identification system (222) analyzes the captured ambient audio to identify one or more associated acoustic features such as mel-frequency, cepstral coefficients, spectral energy, spectral entropy, spectral flatness, and spectral flux. Further, the event identification system (222) compares the identified acoustic features with acoustic features of different audios unrelated to the selected video and pre-stored in the user profile storage system (212). For example, the event identification system (222) compares the identified acoustic features with acoustic features of events such as doorbell sounds, phone ringtones, whistles produced by pressure cookers, oven timer alerts, and human voices. The event identification system (222) then identifies that the captured ambient audio includes, for example, a doorbell sound when the identified acoustic features match with acoustic features of a particular doorbell sound stored in the user profile storage system (212).
[0049] Further, in the previously noted example, the event identification system (222) identifies that an event occurred while the user (112A) was watching the selected video. The event may be identified as “someone pressing a doorbell at the customer premises (104A).” Similarly, the event identification system (222) identifies other events such as “receiving a phone call,” “cooking a meal in a pressure cooker,” “cooking a meal in a microwave oven,” and “conversing with family members” by matching acoustic features of detected sounds with stored features corresponding to phone ringtones, pressure cooker whistles, oven timer alerts, and human voices, respectively stored in the user profile storage system (212). In one embodiment, the event identification system (222) identifies that the user (112A) converses with another family member after pausing the selected video. In this scenario, the event identification system (222) analyzes audio captured by the audio sensor (224) and performs voice recognition to identify if the user (112A) converses with the family member on characters and/or scenes in the selected video. Further, the media player (204) identifies the pausing action of the user (112A) as a positive user action when the event identification system (222) identifies that the user (112A) converses with the family member on the characters and/or scenes in the selected video. Otherwise, the media player (204) identifies the pausing action of the user (112A) as a negative user action when the event identification system (222) identifies that the user (112A) converses with the family member on matters unrelated to the selected video.
[0050] In certain embodiments, the event identification system (222) also identifies events that occurred while the user (112A) is watching the selected video based on one or more images of the viewing area captured by the imaging sensor (208). For example, the imaging sensor (208) captures one or more images of the viewing area and provides the captured images as an input to the event identification system (222). Upon receiving the images, the event identification system (222) processes the images using one or more image processing techniques to identify events such as “entry of a new person into a viewing area”, “exit of a person from the viewing area,” and “usage of a mobile phone while watching the selected video.”
[0051] Upon identifying an occurrence of one or more of the events while the user (112A) is watching the selected video, the media player (204) transmits a set of information including the identified events, emotions detected by the emotion detection system (220), and interactive actions captured by the media player (204) to the content server (106) via the communications link (116). Upon receiving the set of information, the recommendation subsystem (110) determines ratings for different categories embedded as metadata markings in the selected video, different portions of the selected video, and an overall rating for the selected video. Further, the recommendation subsystem (110) generates a preference profile that indicates media content preferences of the user (112A) based on ratings associated with different videos determined by the recommendation subsystem (110), as described subsequently in greater detail with reference to FIGS. 3A-C.
[0052] FIGS. 3A-C illustrate a flow diagram depicting an exemplary method (300) for generating a preference profile that indicates media content preferences of the user (112A) using the recommendation subsystem (110) of FIG. 1. The order in which the exemplary method (300) is described is not intended to be construed as a limitation, and any number of the described blocks may be combined in any order to implement the exemplary method disclosed herein, or an equivalent alternative method. Additionally, certain blocks may be deleted from the exemplary method or augmented by additional blocks with added functionality without departing from the claimed scope of the subject matter described herein.
[0053] At step (302), the media player (204) prompts for registration on the media device (102A) when detecting new users (112A-C) in a viewing area. For example, the media player (204) displays a message “Do you want to register now?” along with ‘Yes’ and ‘No’ options on the display unit (202) of the media device (102A) upon detecting a user (112A) in the viewing area. Upon selecting ‘Yes’ option, the media player (204) enables the imaging-sensor control system (214) and motion-sensor control system (210) to active the camera (208) and the motion sensor (206), respectively. The activated camera (208) captures one or more images of the user’s (112A) face and body in one or more orientations. The image-processing system (218) processes the captured images and identifies facial information of the user (112A) from the captured images, and stores the facial information in the user profile storage system (212). Similarly, the activated motion sensor (206) captures physical attributes of the user (112) such as size, stature, physique, gait information, and other physical information of the user (112A), which are subsequently stored in the user profile storage system (212). Additionally, the media player (204) prompts the user (112A) to select, add, and register one or more events occurring in the real world that may conflict with the viewing activity experience. For example, the media player (204) prompts the user (112A) to initiate events such as pressing the doorbell, ringing of a home or mobile phone, release of a pressure cooker whistle, and an oven timer alert. The audio sensor (224) detects and records the audio corresponding to the events. The media player (204) subsequently stores the recorded audio with the selected event in the user profile storage system (212) for future use.
[0054] At step (304), the motion-sensor control system (210) activates the motion sensor (206) when a user switches on the media device (102A). At step (306), the motion sensor (206) captures physical attributes of the user. At step (308), the motion-sensor control system (210) identifies the user currently located in the viewing area as the user (112A) by comparing the captured physical attributes with physical attributes of various users (112A-C) stored in the user profile storage system (212), as noted previously with reference to FIG. 2.
[0055] At step (310), the motion-sensor control system (210) transmits a control signal to the imaging-sensor control system (214) to activate the camera (208) when the motion-sensor control system (210) fails to identify the user. At step (312), the camera (208) captures one or more images of the viewing area. At step (314), the image processing system (218) identifies the detected user by comparing the facial information obtained from the captured images and facial information of the user (112A) stored in the user profile storage system (212), as noted previously with reference to FIG. 2.
[0056] At step (316), the media player (204) presents a list of one or more videos on the display unit (202) of the media device (102A) for user selection. In one embodiment, the list of one or more videos presented by the media player (204) are a predefined list of videos curated by a content provider or an OTT service provider for users belonging to the same demographic as the registered user detected in the viewing area . In another embodiment, the media player (204) allows the user (112A) to search for one or more videos according to his or her interests, and presents the one or more videos retrieved from the content server (106) for user selection. Further, the media player (204) enables the user (112A) to select a particular video from the one or more videos presented by the media player (204). Subsequently, the media player (204) transmits a request to the content server (106) for streaming the video selected by the user (112A) from the content server (106) to the media player (204). Accordingly, at step (318), the media player (204) receives the selected video along with metadata markings from the content server (106) and queues portions of the selected video for playing. For example, the media player (204) receives a selected video of 7-minutes length along with metadata markings from the content server (106). In one embodiment, the metadata markings include categories associated with different portions of the selected video.

[0057] Table 1 – Metadata Markings For Different Portions of Video

VP 0-30 seconds (s) 30-90 s 30-390 s 390-420 s
MM Opening Credits Adventure Comedy Closing Credits
IW 0.2 1 0.75 0.1
AW 1.77 17.7 79.65 0.88
EM Neutral Surprise Happy Neutral
IA Skip Replay Pause & resume later Terminate play
EV None None Pressing of a doorbell None
ES 0.5 0.8 0.9 0.1
WW 0.8 1.03 1.06 0.6
VPR 0.4 0.824 0.954 0.06
CR 40.11 82.37 95.39 5.68

[0058] In Table 1, ‘VP’ corresponds to different portions of the selected video, “MM” corresponds to metadata markings, ‘IW’ corresponds to initial weightage, ‘AW’ corresponds to actual weightage, ‘EM’ corresponds to emotions, ‘IA’ corresponds to interactive actions, ‘EV’ corresponds to events, ‘and ES’ corresponds to engagement score, ‘WW’ corresponds to watch weightage. Further, ‘VPR’ corresponds to ratings of different video segments and ‘CR’ corresponds to category ratings.
[0059] For example, Table 1 depicts exemplary metadata markings associated with different portions of the selected video. For example, a first portion corresponding to 0-30 seconds of the selected video has metadata markings that correspond to “opening credits.” A second portion corresponding to 30-90 seconds of the selected video has metadata markings that correspond to “adventure.” A third portion corresponding to 30-390 seconds of the selected video has metadata markings that correspond to “comedy.” Further, a fourth portion corresponding to 390-420 seconds of the selected video has metadata markings that correspond to “closing credits.”
[0060] In the previously noted example, at step (320), the media player (204) identifies categories associated with different portions of the selected video from the metadata markings. Specifically, the media player (204) identifies that the first and fourth portions corresponding to 0-30 seconds and 390-420 seconds is related to opening and closing credits, and the second and third portions corresponding to 30-90 seconds and 30-390 seconds is related to adventure and comedy categories, respectively.
[0061] At step (322), the emotion detection system (220) detects emotions of the user (112A) while watching different portions of the selected video. To that end, the emotion detection system (220) includes the camera (208) for capturing one or more images of the user (112A) while the media player (204) is playing different portions of the selected video. For example, the camera (208) captures one or more images of the user (112A) while the media player (204) is playing the first portion of the selected video. Subsequently, the image processing system (218) identifies facial information of the user (112A) from the captured images, and detects an emotion of the user (112A) based on the identified facial information using one or more emotion detection algorithms such as support vector machine, random forest, and Nearest Neighbor Algorithm.
[0062] For example, the image processing system (218) identifies that eyebrows of the user (112A) are at a raised position, upper eyelids of the user (112A) are at a raised position, and a jaw of the user (112A) is dropped to a lowered position from the captured images. In this example, the image processing system (218) detects a prevailing emotion of the user (112A) as “surprise.” In another example, the image processing system (218) identifies that cheeks of the user (112A) are at a raised position and lips of the user (112A) are at a pulled to left and right of the face from the captured images. In this example, the image processing system (218) detects that an emotion of the user (112A) as “happy.” In one embodiment, the image processing system (218) detects that an emotion of the user (112A) as “neutral” when facial features such as eyebrows, eyelids, jaw, cheeks, and lips are at their corresponding neutral positions. The emotion detection system (220) similarly detects emotions of the user (112A) while watching other portions of the selected video from the captured images, as noted previously. Examples emotions of the user (112A) detected by the emotion detection system (220) while watching different portions of the selected video are tabulated in Table 1.
[0063] At step (324), the media player (204) monitors interactive actions performed by the user (112A) during playback of different portions of the selected video. For example, as depicted in Table 1, the media player (204) identifies a first interactive action that includes skipping playback of the first portion. Further, the media player (204) identifies a second interactive action that includes replaying playback of the second portion. In addition, the media player (204) identifies a third interactive action that includes pausing playback of the third portion twice and subsequently resuming playback of the third portion after each instance of pausing the third portion. Furthermore, the media player (204) identifies a fourth interactive action that includes terminating playback of the selected video without completely watching the fourth portion.
[0064] Additionally, at step (326), the event identification system (222) identifies occurrence of one or more events while the user (112A) is watching the selected video. In certain embodiments, the event identification system (222) identifies the events based on inputs received from one or more sensors including the motion sensor (206), the imaging sensor (208), and the audio sensor (224). For example, the audio sensor (224) captures ambient audio generated during playback of the third portion of the selected video. In this example, the event identification system (222) analyzes the captured ambient audio, and further identifies that the captured ambient audio includes both audio related to the selected video and sound of a doorbell deployed at the customer premises (104A). Further, the event identification system (222) identifies occurrence of an event while the user (112A) is watching the third portion as “someone pressing a doorbell at the customer premises (104A).”
[0065] Further, at step (328), the media player (204) determines engagement scores indicating engagement levels of the user (112A) with different portions of the selected video. In one embodiment, the media player (204) determines an engagement score associated with a particular video portion based on a total duration of the video portion and a duration of the video portion watched by the user (112A), for example, using equation (1):

ES=WD/TD (1)

where, ‘ES’ corresponds to an engagement score of the video portion, ‘WD’ corresponds to a duration of the video portion watched by the user (112A), and ‘TD’ corresponds to a total duration of the video portion.

[0066] In certain embodiments, the image processing system (218) computes a duration of the video portion watched by the user (112A) by identifying eye gazing direction of the user (112A) from images of the user (112A) captured using the camera (208). For example, the image processing system (218) analyzes eye gazing direction of the user (112A) from the captured images, and identifies that the user (112A) has gazed only at the media device (102A) during playback of the first portion that is of 30-seconds length. Further, the media player (204) identifies that the user (112A) has watched only first 15-seconds of the first portion and skipped rest 15-seconds of the first portion. In this example, the image processing system (218) computes that a duration of the first portion watched by the user (112A) as 15-seconds. Accordingly, the media player (204) identifies an engagement score for the first portion as 0.5 using equation (1).
[0067] In another example, the image processing system (218) analyzes eye gazing direction of the user (112A), and identifies that the user (112A) has gazed at the media device (102A) only for 48-seconds during playback of the second portion that is of 60-seconds length. Further, the image processing system (218) identifies that the user (112A) was involved in using his/her mobile phone during playback of remaining 12-seconds of the second portion. In this example, the image processing system (218) computes that a duration of the second portion watched by the user (112A) as 48-seconds. Accordingly, the media player (204) identifies an engagement score for the second portion as 0.8.
[0068] In yet another example, the image processing system (218) analyzes eye gazing direction of the user (112A), and identifies that the user (112A) has gazed at the media device (102A) only for 324-seconds during playback of the third portion that is of 360-seconds length. Further, the image processing system (218) identifies that the user (112A) was involved in conversing with his or her family member during playback of remaining 36-seconds of the third portion. In this example, the image processing system (218) computes that a duration of the third portion watched by the user (112A) as 324-seconds. Accordingly, the media player (204) identifies an engagement score for the third portion as 0.9 in accordance with the previously noted equation (1). Further, the media player (204) similarly identifies an engagement score for the fourth portion as 0.1 when a total duration of the fourth portion is 30-seconds and a duration of the fourth portion watched by the user (112A) is 3-seconds.
[0069] In certain embodiments, instead of determining engagement scores, the media player (204) transmits duration information including a duration for which each of the video portions is watched by the user (112A) to the content server (106). Subsequently, the recommendation subsystem (110) in the content server (106) determines an engagement score associated with each of the video portions, using equation (1).
[0070] Subsequently, at step (330), the media player (204) transmits a set of information including the detected emotions, the identified interactive actions, the identified events, and the determined engagement scores to the content server (106) via the communications link (116). The recommendation subsystem (110) uses the set of information received from the media player (204) to determine ratings for different categories associated with the selected video, different portions of the selected video, and an overall rating for the selected video.
[0071] Specifically, at step (332), the recommendation subsystem (110) assigns initial weightages for different portions of the selected video based on categories of those video portions and certain predefined rules. For example, a predefined rule may define initial weightages to be assigned to video portions having opening credits related content, adventure related content, comedy related content, and closing credits related content as 0.2, 1, 0.75, and 0.1, respectively. Accordingly, as tabulated in Table 1, the recommendation subsystem (110) assigns initial weightages for the first, second, third, and fourth portions of the selected video as 0.2, 1, 0.75, and 0.1, respectively.
[0072] At step (334), the recommendation subsystem (110) determines an average weightage for each portion of the selected video. In one embodiment, the recommendation subsystem (110) determines the average weightage for a particular video portion, for example, using equation (2):

AVGW=(TD*(1*IW))/TIW (2)

where, ‘AVGW’ corresponds to an average weightage of a particular video portion, ‘TD’ corresponds to a total duration of the video portion, ‘IW’ corresponds to an initial weightage of the video portion, and ‘TIW’ corresponds to a total of initial weightages assigned to different portions of the selected video.

[0073] For example, the recommendation subsystem (110) determines an average weightage for the first portion as 2.93 when the total duration of the first portion corresponds to 30-seconds, the initial weightage of the first portion corresponds to 0.2, and the total of initial weightages assigned to different portions of the selected video corresponds to 2.05. Similarly, as tabulated in Table 1, the recommendation subsystem (110) determines average weightages of the second, third, and fourth portions of the selected video as 29.27, 131.71, and 1.46, respectively using equation (2).
[0074] At step (336), the recommendation subsystem (110) determines an actual weightage for each portion of the selected video. In one embodiment, the recommendation subsystem (110) determines an actual weightage of a particular video portion, for example, using equation (3):

AW=(AVGW*100)/TAVGW (3)

where, ‘AW’ corresponds to an actual weightage of a particular video portion, ‘AVGW’ corresponds to an average weightage of the particular video portion, and ‘TAVGW’ corresponds to a total of average weightages determined for different portions of the selected video.

[0075] For example, the recommendation subsystem (110) determines an actual weightage of the first portion as 1.77 when an average weightage of the first portion corresponds to 2.93, and a total of average weightages determined for different portions of the selected video corresponds to 165.37, using equation (3). The recommendation subsystem (110) similarly determines actual weightages of the second, third, and fourth portions of the selected video as 17.7, 79.65, and 0.88, respectively using equation (3) and as tabulated in Table 1.
[0076] Further, at step (338), the recommendation subsystem (110) determines a watch weightage for each portion of the selected video. In one embodiment, the watch weightage determined for a particular portion of the selected video indicates a level of interest expressed by the user (112A) on the particular video portion. The level of interest, for example, may be determined based on emotions of the user while watching the particular video portion, a duration for which the user (112A) gazed at the media device (102A) while watching the particular video portion, and one or more interactive actions performed by the user while watching the particular video portion. In certain embodiments, the recommendation subsystem (110) determines watch weightages for video portions including opening and closing credits based on default values stored in a weightage database (118) of the recommendation subsystem (110). For example, a default value stored in the weightage database (118) for a video portion containing opening credits is 0.8 and another default value stored in the weightage database (118) for a video portion containing closing credits is 0.6. In the previously noted example, the recommendation subsystem (110) determines watch weightages of the first and fourth portions including opening and closing credits as 0.8 and 0.6, respectively based on their corresponding default values stored in the weightage database (118).
[0077] In certain embodiments, the recommendation subsystem (110) determines watch weightages for other portions, for example, the second and third portions of the selected video and that lack content related to opening and closing credits, for example, using equation (4).

WW=(DV+IASW) (4)

where, ‘WW’ corresponds to a watch weightage of a particular video portion, ‘DV’ corresponds to a default value stored in the weightage database (118) for a video portion that is not related to opening and closing credits, and ‘IASW’ corresponds to an interactive action specific weightage.

[0078] For example, the recommendation subsystem (110) determines a watch weightage for the second portion that belongs to adventure category based on the default value, an interactive action performed by the user (112A) while watching the second portion, and a weightage related to the interactive action performed by the user (112A). For instance, the recommendation subsystem (110) identifies that the user (112A) has replayed the second portion after watching it completely from the set of information received from the media player (204). In this example, the recommendation subsystem (110) determines a watch weightage for the second portion as 1.03 when the default value corresponds to 1 and a weightage related to a video portion that is replayed corresponds to 0.3, using equation (4).
[0079] Similarly, the recommendation subsystem (110) determines a watch weightage for the third portion corresponding to comedy category based on the default value, an interactive action performed by the user (112A) while watching the third portion, and a weightage related to the interactive action performed by the user (112A). In one embodiment, the recommendation system determines the weightage related to the interactive action performed by the user (112A) based on identification of occurrence of an event within a particular time period prior to the interactive action performed by the user (112A). For example, the recommendation subsystem (110) identifies that the user (112A) has paused playback of the third portion twice, for example, at 10:00:05 AM for the first time and at 10:12:06 AM for the second time. Further, from the events information received from the media player (204), the recommendation subsystem (110) identifies that a first event occurred at 10:00:00 AM, that is, five seconds before the first pausing instance and a second event occurred at 10:12:00 AM, that is, six second before the second pausing instance. An example of the first event includes identification of sound of a doorbell detected at the premises (104A) of the user (112A)”. An example of the second event includes identification of ringing of a phone.
[0080] In the previously noted examples, the recommendation subsystem (110) identifies that the user (112A) has paused the third portion for the first time to attend a person waiting outside the customer premises (104A). Similarly, the recommendation subsystem (110) identifies that the user (112A) has paused the third portion for the second time to attend a phone call received by the user (112A). Accordingly, the recommendation subsystem (110) identifies that the user (112A) has paused the third portion (112A) in both instances to perform certain tasks and not because of his or her disinterest in the selected video. Further, in this example, the recommendation subsystem (110) does not identify a pausing action of the third portion automatically as a negative user action. Instead, the recommendation subsystem (110) ignores the pausing action when the user (112A) resumes playback of the paused third portion within a particular time period, for example, within a few minutes or within a week’s time. Additionally, the recommendation subsystem (110) assigns a first weightage of +0.03 for each instance of pausing and subsequently resuming playback of the third portion. Accordingly, the recommendation subsystem (110) determines the watch weightage of the third portion as 1.06 as tabulated in Table 1 when the default value corresponds to 1, and a weightage relating to each instance of pausing and subsequently resuming playback of the video portion corresponds to +0.03, using equation (4).
[0081] In another embodiment, the recommendation subsystem (110) identifies pausing of a video as a negative interactive action when the recommendation subsystem (110) identifies that no events occurred within a particular time period, for example within ten seconds, prior to pausing of the video. For example, the recommendation subsystem (110) identifies that the user (112A) paused playback of the third portion once at 10:00:05 AM. Further, the recommendation subsystem (110) identifies that no events occurred in last ten seconds prior to pausing of the third portion. In this example, the recommendation subsystem (110) identifies that the user (112A) has paused the third portion because of his or her disinterest in the selected video, and accordingly, assigns a second weightage of -0.03 for pausing playback of the third portion. Further, in this example, the recommendation subsystem (110) determines the watch weightage of the third portion as 0.97 when the default value corresponds to 1, and a weightage relating to pausing the video portion corresponds to -0.03, using equation (4).
[0082] Furthermore, at step (340), the recommendation subsystem (110) determines ratings for different portions of the selected video. For example, the recommendation subsystem (110) determines a rating for a portion of the selected video, for example, using equation (5):

VPR=ES*WW (5)

where, ‘VPR’ corresponds to a rating of a particular video portion, ‘ES’ corresponds to an engagement score, and ‘WW’ corresponds to a watch weightage.

[0083] For example, the recommendation subsystem (110) determines a rating of the first portion as 0.4 when the engagement score and watch weightage of the first portion are 0.5 and 0.8, respectively. Similarly, the recommendation subsystem (110) determines ratings for the second, third, and fourth portions of the selected video as 0.824, 0.954, and 0.06, respectively, using equation (5) and as tabulated in Table 1.
[0084] At step (342), the recommendation subsystem (110) determines ratings for different categories associated with the selected video. In one embodiment, the recommendation subsystem (110) determines a rating for a particular category associated with the selected video, for example, using equations (6) and (7):

IR=VPR*AW (6)

CR=(IR*100)/AW (7)

‘where, ‘IR’ corresponds to an initial rating associated with a particular category of the selected video, ‘CR’ corresponds to a category rating associated with the particular category, ‘VPR’ corresponds to a rating of a video portion containing the particular category of content, and ’AW’ corresponds to an actual weightage of the particular video portion.

[0085] For example, the recommendation subsystem (110) determines an initial rating of the category “opening credits” as 0.71 when the rating of the first portion and the actual weightage of the first portion corresponds to 0.4 and 1.77, respectively, using equation (6). The recommendation subsystem (110) further determines a rating for the category “opening credits” as 40.11 when the initial rating corresponds to 0.71 and the actual weightage of the first portion corresponds to 1.77, using equation (7). Similarly, the recommendation subsystem (110) determines initial weightages of other categories of the selected video including “adventure”, “comedy”, and “closing credits” as 14.58, 75.98, and 0.05, respectively, using equation (6). Further, the recommendation subsystem (110) determines ratings of the categories “adventure”, “comedy”, and “closing credits” as 82.37, 95.39, and 5.68, respectively, using equation (7).
[0086] Following determining of category ratings of various portions of the selected video, at step (344), the recommendation subsystem (110) determines an overall rating for the selected video. In one embodiment, the recommendation subsystem (110) determines the overall rating by adding ratings determined for individual portions of the selected video. For example, the recommendation subsystem (110) determines that the overall rating of the selected video as 2.238 when ratings associated with the first, second, third, and fourth video portions correspond to 0.4, 0.824, 0.954, and 0.06, respectively.
[0087] At step (346), the recommendation subsystem (110) classifies categories associated with the selected video into one of a highly preferred category, a moderately preferred category, and a less preferred category based on associated ratings. In one embodiment, the recommendation subsystem (110) classifies a category associated with the selected video into highly, moderately, or less preferred category when a rating of the category is above 70, in between 40 to 70, or less than 40, respectively. For example, the recommendation subsystem (110) classifies both categories such as ‘adventure’ and ‘comedy’ associated with the selected video into a highly preferred category as corresponding ratings are 82.37 and 95.39.
[0088] At step (348), the recommendation subsystem (110) generates a preference profile for the user (112A) based on ratings determined for the categories of the selected video, and stores the generated preference profile in the database (114). In one embodiment, the generated preference profile includes categories such as ‘adventure’ and ‘comedy’ classified under the highly preferred category, and ratings determined for ‘adventure’ and ‘comedy’ categories. Further, as the user (112A) watches more videos, the recommendation subsystem (110) determines ratings for various categories associated with those videos based on the emotions, interactive actions, events, and engagement score information received from the media player (204) associated with the media content recommendation system (100). Further, the recommendation subsystem (110) updates the preference profile of the user (112A) by including all those categories under one of the highly preferred, moderately preferred, and less preferred category along with associated ratings.
[0089] Though FIGS. 3A-C depict and describe an exemplary methodology for generating a preference profile for the single user (112A), it is to be understood that the recommendation subsystem (110) similarly generates preference profiles for other users (112B-C) in the customer premises (104A) based on their corresponding emotions, interactive actions, events, and engagement scores information received from the media player (204). Likewise, the recommendation subsystem (110) also generates preference profiles for users (112D-N) in other premises (104B-N).
[0090] In certain embodiments, the recommendation subsystem (110) provides content recommendations that are significantly more accurate than content recommendations provided by conventional recommendation systems, as depicted and described subsequently with reference to Table 2 and Table 3.

[0091] Table 2 – Ratings Determined By The Recommendation Subsystem (110)

VP 0-30 seconds (s) 30-90 s 30-390 s 390-420 s
MM Opening Credits Adventure Comedy Closing Credits
IW 0.2 1 0.75 0.1
AW 1.77 17.7 79.65 0.88
EM Neutral Surprise Happy Neutral
IA Skip Replay Pause 10 times & resume later Terminate play
EV None None None None
ES 0.5 0.8 0.9 0.1
WW 0.8 1.03 0.7 0.6
VPR 0.4 0.824 0.63 0.06
CR 40.11 82.37 63 5.68

[0092] For example, as tabulated in Table 2, the user (112A) pauses the selected video ten different times while watching the third portion and subsequently resumes playback after each instance of pausing the third portion. In this example, the recommendation subsystem (110) also identifies that no event occurred prior to each instance of pausing the playback of third portion based on events information received from the media player (204).
[0093] In the previously noted example, the recommendation subsystem (110) identifies each instance of pausing the playback of third portion as a negative input. In an exemplary scenario, the recommendation subsystem (110) identifies that the user (112A) has paused the third portion and subsequently watched other videos before resuming the playback of third portion based on a set of information received from the media player (204). This user behavior of switching to other videos without completely watching the selected video is determined by the recommendation subsystem (110) to indicate that the user (112A) is less interested in the selected video. Accordingly, the recommendation subsystem (110) assigns a negative weightage of -0.03 for each instance of pausing the playback of third portion. Further, as tabulated in Table 2, the recommendation subsystem (110) determines the watch weightage of the third portion as 0.7 when the default value corresponds to 1, and when a negative weightage relating to each instance of pausing the third portion corresponds to -0.03, using equation (4).
[0094] Further, as tabulated in Table 2, the recommendation subsystem (110) determines a rating for the third portion as 0.63 using equation (5), initial rating for the third portion as 50.18 using equation (6), and rating for the category “comedy” associated with the third portion as 63 using equation (7). Further, in this example, the recommendation subsystem (110) generates a preference profile of the user (112A) including the category “comedy” classified under a moderately preferred category. Post generating the preference profile, the recommendation subsystem (110) will only occasionally, for example twice a month, recommend comedy related content to the user (112A) as the comedy category is only moderately preferred by the user (112A).
[0095] However, in the same example scenario described previously, a conventional recommendation system may determine a rating for the category “comedy” as 71.685.

[0096] Table 3 – Ratings Determined By Conventional Recommendation System

VP 0-30 seconds (s) 30-90 s 30-390 s 390-420 s
MM Opening Credits Adventure Comedy Closing Credits
IW 0.2 1 0.75 0.1
AW 1.77 17.7 79.65 0.88
EM Neutral Surprise Happy Neutral
IA Skip Replay Pause 10 times & resume later Terminate play
EV None None None None
UIS 0.5 0.8 0.9 0.1
Ratings 0.885 14.16 71.685 0.088

[0097] In Table 3, ‘UIS’ corresponds to a user interest score, which is different from an engagement score determined by the recommendation subsystem (110). For example, the conventional recommendation system determines a user interest score for a video that is of 300-seconds length as ‘1’ or ‘100%’ when a media player plays all 300-seconds of the video. However, allowing playback of the video for the entire duration does not necessarily indicate that the user is interested on the video and has watched the video completely. In the real world scenario, the user may actually find the video as not interesting and start engaging in other activities such as browsing on a phone when the media player continues to play the video. In such a scenario, the conventional recommendation system does not identify that the user is not interested on the video and merely identifies that the user interest level is 100% if the video is played completely. However, in the same exemplary scenario, the recommendation subsystem (110) identifies an exact duration of the video watched by the user by analyzing eye gazing direction of the user, and determines an engagement score based on the identified duration. For example, the user has watched first 120 seconds of the video and started browsing the phone after that. In this example, the recommendation subsystem (110) accurately determines the engagement score for the video as 0.4 in contrast to the user interest score of ‘1’ determined by the conventional recommendation system. Thus, the recommendation subsystem (110) of the present disclosure provides more accurate recommendations by accurately identifying engagement scores for different videos.
[0098] Specifically, as tabulated in Table 3, the conventional recommendation system determines the rating for the third portion and for the category “comedy” associated with the third portion as 71.685 by multiplying the user interest score of 0.9 with the actual weightage of 79.65 associated with the third portion of the selected video. This is because the conventional recommendation system does not generally identify occurrence of an event at a designated time period prior to pausing of the video, and identify the context behind pausing playback of the video. The conventional recommendation system merely identifies all pausing activities of the user (112A) as a negative input. However, in real-world scenarios, the user (112A) may need to pause a video for several reasons even when the user (112A) likes the video. For example, the user (112A) may pause the video to answer a phone call received by the user (112A), to turn off a cooking device, to attend a person waiting outside the home, or when a neighbor visits the home. The conventional recommendation system does not identify these circumstances that force the user (112A) to pause the video, and identifies all pausing activities of the user (112A) as a negative input, which leads to inaccurate identification of media content preferences of the user (112A), thus resulting in inaccurate recommendations.
[0099] Further, in the previously noted example, the conventional recommendation system identifies the category “comedy” as highly preferred by the user (112A) as the associated rating of 71.685 is above 70. However, in reality, the category “comedy” is only moderately preferred by the user (112A) as accurately identified by the recommendation subsystem (110) of FIG. 1. Thus, the recommendation subsystem (110) identifies content preferences of the user (112A) more accurately when compared to conventional recommendation systems, thereby providing more accurate content recommendations to the user (112A). This leads to greater user satisfaction, while conserving bandwidth and other network resources typically wasted during browsing through and playing samples of multiple videos by the user (112A) when trying to select something of his or her interest to play.
[0100] In addition to identifying individual content preferences of the user (112A), and other users (112B-C) in the customer premises (104A), the recommendation subsystem (110) also provides recommendation including a set of media content of interest to the combined group of users (112A-C). An exemplary method for providing content recommendations to a group of users using the media content recommendation system (100) is described in greater detail with reference to FIGS. 4A-B.
[0101] FIGS. 4A-B illustrate a flow diagram depicting an exemplary method (400) for providing content recommendation to two or more users (112A-C) watching a video individually, or together as a group, using the recommendation subsystem (110) of FIG. 1. Specifically, at step (402), the ambient sensors (203) identify one or more users currently watching a video presented by the media device (102A). For example, one or more of the motion-sensor control system (210) or the image processing system (218) identifies that only one user (112A) is currently watching the video presented by the media device (102A). In this example, the recommendation subsystem (110) identifies categories of content that are highly preferred by the identified user (112A) from his or her preference profile. The recommendation subsystem (110) then recommends a set of videos to the user (112A), where the set of videos belong to the categories that are highly preferred by the user (112A).
[0102] Alternatively, the motion-sensor control system (210) or the image processing system (218), for example, identifies that there is a group of three users (112A-C) belonging to a family currently watching the video presented by the media device (102A). Subsequently, at step (404), the recommendation subsystem (110) retrieves preference profiles of the identified users (112A-C) from the preference database (114). For example, Table 4 depicts exemplary preference profiles of the users (112A-C) and that are retrieved from the preference database (114).

[0103] Table 4 – Preference Profiles of Users

Content Preferences Preference Profile 1 Preference Profile 2 Preference Profile 3
Categories Ratings Categories Ratings Categories Ratings
Highly Preferred Adventure 82.37 Comedy 90 Sports 87
Comedy 95.39 Sports 85 Comedy 81
Moderately Preferred Travel 64 Science 68 Animation 55
Food 61 Fantasy 62 Romance 40
Less Preferred Historic 30 Action 25 Drama 30
Horror 20 Politics 15 Thriller 28

[0104] In Table 4, the ‘preference profiles 1’, ‘preference profile 2’, and ‘preference profile 3’ indicate preference profiles of the first user (112A), second user (112B), and third user (112C), respectively. Referring back to FIG. 4A, at step (406), the recommendation subsystem (110) identifies one or more content categories that are highly and commonly preferred by the users (112A-C) from the retrieved preference profiles. For example, from the preference profiles retrieved from the preference database (114), the recommendation subsystem (110) identifies that a content category that is highly and commonly preferred by all the users (112A-C) is “comedy.” Subsequently, the recommendation subsystem (110) recommends a list of media content belonging to a comedy genre to the users (112A-C). In one embodiment, the recommended list includes comedy programs that have not been previously watched by any of the users (112A-C).
[0105] In another example, the motion-sensor control system (210) or the image processing system (218), for example, identifies that there are only two users (112B-C) currently watching the video presented by the media device (102A). In this example, the recommendation subsystem (110) identifies that a content category that is highly and commonly preferred by the users (112B-C) is “sports” from the preference profiles 2 and 3 retrieved from the preference database (114). Subsequently, the recommendation subsystem (110) recommends a list of videos to the users (112B-C), where the recommended videos are related to the category “sports.” Thus, the present recommendation subsystem (100) identifies and monitors a group of individual users who are all currently watching the video presented by the media device (102A) in near real-time, and dynamically tailors more accurate recommendations to the group according to content preferences of those individual users.
[0106] However, conventional recommendation systems do not generally identify and track individual users in the group and recommend media content that satisfy content preferences of all individual users in the group. The conventional recommendation systems provide only static recommendations based on watch histories of a user who has originally logged into an account with the media player (204), where such static recommendations may be irrelevant to all or some of the users who are currently watching media content presented by the media device (102A). Thus, the present recommendation subsystem (110) provides more accurate media content recommendations when compared to conventional recommendation systems.
[0107] At step (408), the recommendation subsystem (110) identifies a list of videos to be transmitted to the media device (102A) as recommendations. In one embodiment, the recommendation subsystem (110) identifies the list of videos based on content categories that are highly and commonly preferred by the users (112A-C), where the list of identified videos are also rated high by other groups having content preferences similar to content preferences of the users (112A-C) group. For example, the recommendation subsystem (110) identifies a list of videos related to the category ’comedy’ that is highly and commonly preferred by the users (112A-C). Further, the list of videos identified by the recommendation subsystem (110) includes comedy videos rated high by other similar demographic groups having content preferences similar to content preferences of the users (112A-C) group.
[0108] In certain embodiments, the recommendation subsystem (110) may be implemented as a machine learning system that learns watch patterns of the users (112A-C) in the group. For example, when the users (112-C) watch media content together as a group, the recommendation subsystem (110) learns that videos are consistently selected according to preferences of a particular user (112A) in the group. In this example, the recommendation subsystem (110) identifies a list of videos to be transmitted to the media device (102A) as recommendations based on preferences of the particular user (112A) in the group. Additionally, the recommendation subsystem (110) may also identify a list of videos to be transmitted to the media device (102A) as recommendations based on preferences of other users in similar demographic groups as the users (112A-C).
[0109] At step (410), the content server (106) transmits the list of videos identified by the recommendation subsystem (110) to the media device (102A) via the communications link (116). At step (412), the media player (204) plays a particular video selected by the user (112A) from the list of videos provided as the recommendation. At step (414), the recommendation subsystem (110) determines multi-level ratings for each of the users (112A-C) including a category rating, individual portions ratings, and an overall rating for the selected video. In one embodiment, the recommendation subsystem (110) determines the multi-level ratings for each of the users (112A-C) using equations (1)-(7), as noted previously with reference to FIGS. 3A-C. It may be noted that the multi-level ratings determined for one user may not be same as the multi-level ratings determined for another user as each user may have different engagement level and different emotions while watching different portions of the selected video.
[0110] At step (416), the recommendation subsystem (110) updates the preference profiles of the users (112A-C) based on their corresponding multi-level ratings determined by the recommendation subsystem (110). For example, the recommendation subsystem (110) determines that ratings of a category ’comedy’ associated with the selected video are 97, 91, and 85 for the first, second, and third users (112A-C), respectively. In this example, the recommendation subsystem (110) updates a current rating of 95.39 for the category ‘comedy’ included in the ‘preference profile 1’ to 96.2 by determining an average of the current rating and the category rating determined for the first user (112A). Similarly, the recommendation subsystem (110) updates current ratings of 90 and 81 for the category ‘comedy’ included in the ‘preference profiles 2 and 3’ to 90.5 and 83, respectively by averaging the corresponding current ratings, and ratings determined for the second and third users (112B-C).
[0111] At step (418), the recommendation subsystem (110) determines multi-level ratings for a group including the users (112A-C). In one embodiment, the recommendation subsystem (110) determines the multi-level ratings for the group by averaging the multi-level ratings determined for individual users (112A-C). For example, the recommendation subsystem (110) determines a group rating for the category ‘comedy’ as 91 by averaging individual ratings 97, 91, and 85 determined by the recommendation subsystem (110) for the first, second, and third users (112A-C), respectively.
[0112] At step (420), the recommendation subsystem (110) updates the preference database (114) based on the multi-level ratings determined for the group of users (112A-C). For example, the recommendation subsystem (110) updates the preference database (114) with the group rating of 91 determined for the category ’comedy’ and by classifying the category ’comedy’ under highly preferred categories of the group. As the users (112A-C) watch more videos as a group, the recommendation subsystem (110) similarly determines group ratings for various categories associated with different videos and updates the preference database (114) with the determined group ratings. The recommendation subsystem (110) further identifies categories of content that are highly and commonly preferred by all the users (112A-C) in the group from the updated preference database (114), to provide future recommendations to the users (112A-C) including a set of media content belonging to the identified categories.
[0113] The recommendation subsystem (110) described in the present disclosure, thus, identifies the context behind interactive actions performed by a user to determine if the interactive actions have to be considered as positive inputs or negative inputs. For example, if a user pauses a video and exits a viewing area to attend a person waiting outside the home, conventional recommendation systems identify the action to indicate lack of user interest in the video, which may not be accurate. Further, the conventional recommendation systems identify pausing playback of the video as a negative input and may not provide future recommendations including videos that are similar to the paused video.
[0114] In contrast, the present recommendation subsystem (110) correctly identifies that the context behind pausing playback of the video is to attend a person waiting outside the home. Accordingly, in the previously noted example, the recommendation subsystem (110) identifies that pausing and subsequently resuming playback of the video as a positive input and provides future recommendations including videos that are similar to the paused and subsequently resumed video. Hence, the recommendation subsystem (110) provides content recommendations that are more accurate when compared to content recommendations provided by conventional recommendation systems.
[0115] Further, as noted previously, conventional recommendation systems rely on user actions such as providing ratings to identify users’ interest levels on media content that they have watched to provide subsequent content recommendations. However, most users such as elderly users and kids do not voluntarily rate the media content they watch, and therefore, content ratings usually reflect the preferences of only a small subset of technically aware users. The present recommendation subsystem (110) does not require users to actively rate media content that they watched. The recommendation subsystem (110) automatically determines ratings and captures interest levels of users related to different media content by analyzing their emotions, interactive actions, occurrence of events within a particular time period prior to interactive actions performed by the users, engagement levels of the users while watching those media content. Therefore, the recommendation subsystem (110) is capable of providing more accurate content recommendations even when the users do not voluntarily rate media content they watch.
[0116] Since the recommendation subsystem (110) itself provides more accurate content recommendation to the user (112A), the user (112A) does not need to manually search for media content that suits his or her preferences. Thus, the recommendation subsystem (110) enhances user’s satisfaction, saves user’s time typically wasted in identifying appropriate media content, saves central processing unit (CPU) utilization typically involved in manual searching of appropriate media content, and saves bandwidth consumption.
[0117] Although specific features of various embodiments of the present systems and methods may be shown in and/or described with respect to some drawings and not in others, this is for convenience only. It is to be understood that the described features, structures, and/or characteristics may be combined and/or used interchangeably in any suitable manner in the various embodiments shown in the different figures.
[0118] While only certain features of the present systems and methods have been illustrated and described herein, many modifications and changes will occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes.

LIST OF NUMERAL REFERENCES:

100 Media Content Recommendation System
102A-N Media Devices
104A-N Customer Premises
106 Content Server
108 Content Database
110 Recommendation Subsystem
112A-N Users
114 Preference Database
116 Communications Link
118 Weightage Database
202 Display Unit
203 Ambient Sensors
204 Media Player
206 Motion Sensor
208 Imaging Sensor
210 Motion-sensor Control System
212 User Profile Storage System
214 Imaging-sensor Control System
218 Image Processing System
220 Emotion Detection System
222 Event Identification System
224 Audio Sensor
300-348 Steps of method for generating a preference profile indicating media content preferences of a user
400-420 Steps of method for providing content recommendation to users watching a video together as a group