Audio-video equipment intelligent management method and system based on multi-source signals

Through the intelligent management method of multi-source signal audio-visual equipment, deep learning and support vector machine technology are used to analyze user viewing data and emotions, and automatically update video recommendations, solving the problem of single video types caused by personalized recommendations and improving user experience Diversity and loyalty.

CN119052537AInactive Publication Date: 2024-11-29深圳市大屏影音技术有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411531579.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-30
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Personalized video recommendations in existing audio and video equipment have led to the reduction of the variety and novelty of the types of videos that users watch, forming an "information cocoon" and losing the novelty and diversity of user experience.

Method used

Using intelligent management methods of audio and video equipment based on multi-source signals, a video classifier is built through deep learning models and support vector institutions to generate video databases, and combined with user viewing data analysis, it is divided into hobby groups, pending groups and blocking groups, and real-time adjustments are made. Video recommendations, automatically update the video types according to user emotions and viewing habits.

Benefits of technology

It improves the comfort and loyalty of users to watch videos, increases the diversity and novelty of videos, reduces the burden of user decision-making, improves the depth and accuracy of personalized recommendations, and creates a more considerate user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119052537A_ABST
    Figure CN119052537A_ABST
Patent Text Reader

Abstract

The invention discloses an intelligent management method and system for audio-video equipment based on multi-source signals, and relates to the technical field of personalized recommendations, and the method comprises the steps: classifying all collected historical videos by using a deep learning model to generate a video database, and training the classified videos by using a support vector machine to construct a video classifier; carrying out secondary division on all video types in the video database, and respectively constructing a hobby group, a shielding group and an undetermined group; marking the watching time of the video type users in the hobby group to generate a user habit list; calculating to obtain a video updating period and an emotion normal threshold value; when the user watches the video by using the audio-video equipment, acquiring the time when the user uses the audio-video equipment in real time, and automatically playing the video in the hobby group of the user; before the video in the user hobby group is automatically played, the time difference value between the real-time time and the last user emotion abnormity is calculated, and whether the time difference value reaches the video updating period is judged.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of personalized recommendation technology, and in particular to an intelligent management method and system for audio-visual equipment based on multi-source signals. Background Art

[0002] With the advancement of technology, traditional analog videos are gradually replaced by digital videos. Digitalization makes video storage, transmission and processing more efficient; the development of cloud technology makes the storage and processing of video data no longer dependent on local hardware, and users can easily manage and share video content on the cloud platform; with the maturity of algorithms and AI technology, video management systems gradually introduce machine learning algorithms to achieve personalized recommendations, video classification and content analysis, and improve user experience; with the further development of technology, the video management of audio and video equipment in the future will be more intelligent, and can provide more accurate recommendations and more personalized user experience. Although as video recommendations in audio and video equipment tend to be more personalized, on the one hand, it greatly fits the user's preferences and improves the user experience; but on the other hand, due to personalized video content recommendations, the diversity of the types of videos watched by users is greatly squeezed, so that users can only see what they want to see, and cannot provide content that users do not know; it constitutes an alternative "information cocoon", which makes the videos watched by users lose novelty and diversity. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent management method and system for audio-visual equipment based on multi-source signals to solve the problems raised in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solutions: An intelligent management method for audio-visual equipment based on multi-source signals, the method comprising the following steps: S100, collecting all videos played by audio and video devices in history, classifying all the collected videos in history using a deep learning model to generate a video database, and training the classified videos using a support vector machine to construct a video classifier; Furthermore, the specific steps of using the support vector machine to train the classified videos and construct a video classifier are as follows: S101. Collect all videos played by audio and video devices in history, and use convolutional neural network to extract features of each frame of each collected video. The formula is: ; In the formula, Represents the features of each frame of each video extracted, CNN represents the convolutional neural network algorithm, Represents each frame of each video; The recurrent neural network algorithm is used to combine the extracted features of each frame with the time sequence to perform time series modeling, and the time sequence of each frame in each video in the history is extracted. The formula is: ;In the formula, represents the constructed time series model, LSTM represents the recurrent neural network, Represents the characteristics of each frame, Indicates the time sequence corresponding to each frame; The deep learning model is constructed by combining the feature extraction and timing model of each frame. The generated deep learning model is In the deep learning model, out represents the output result, softmax represents the activation function, and Dense represents the fully connected layer. The constructed deep learning model is used to classify all the collected videos, and the classification results are , represents the first, second, third, ..., nth video types obtained by classification, where n is a positive integer; a video database is generated using the classified video categories; S102. After classifying the videos in history, the classified video database is input into the support vector machine as a training set to train a video classifier; the audio-visual equipment uses the wireless network to advance the video in the cloud, and after classifying the extracted video using the video classifier, it is transmitted to the video category corresponding to the video database.

[0005] Through deep learning models, all data played in audio-visual devices in history are classified to generate a video database, and the video database is used to train the support vector machine to build a video classifier. After the audio-visual device extracts the video from the network, the classifier is used to perform real-time classification. After the video is classified, users can quickly find videos of interest according to different classifications when watching videos.

[0006] S200, collecting viewing data of all videos of audio-visual device users in history, analyzing viewing data of all videos in user history, dividing all video categories in the video database into secondary categories, respectively constructing a favorite group, a blocked group, and a pending group; marking viewing time of users of video categories in the favorite group, and generating a user habit list; Furthermore, the specific steps for generating a user habit list are: S201, collect the viewing data of all videos by audio-visual device users in history, and extract the viewing times of each type of video in the viewing data. , Indicates the number of times the user has viewed the 1st, 2nd, 3rd, ..., nth type of video in history; after sorting the number of views from small to large, calculate the difference in the number of views of adjacent types of videos, and the difference in all the number of views is calculated. , It means that the differences of the 1st, 2nd, 3rd, ... n-1th adjacent viewing times are calculated, the n-1 viewing times differences are judged, the two difference generation nodes with the largest viewing times differences are selected, the first largest difference in the sorting is used as the node 1 of the shielded group and the pending group, and the second largest difference in the sorting is used as the node 2 of the shielded group and the favorite group; all video types before node 1 in the sorting are used as the shielded group; the video types between node 1 and node 2 in the sorting are used as the pending group; the video types after node 2 in the sorting are used as the favorite group; S201, extracting the viewing time of the video types in the generated hobby group in the history of the user, and marking each video type in the hobby group with the viewing time , Indicates the 1st, 2nd, 3rd, ..., mth video types in the favorite group after marking; generates a user habit list using the marked video types and time.

[0007] By using the viewing data of all types of videos in the user's history, we can analyze the user's preference for each type of video, and divide all videos into a favorite group, a pending group and a blocked group. The blocked group represents the types of videos that the user does not like, and the audio-visual equipment will block such videos and not play them; and the time the user watches videos in the favorite group is recorded to generate a user habit list; when the user uses the audio-visual equipment, personalized and humanized video recommendations are provided based on the user habit list, which greatly improves the user's comfort in watching videos.

[0008] S300, according to the user habit list record history, the user watches a video of a type not recorded in the user habit list at each habit time, the event is judged as abnormal user emotion, and the time difference of all records is counted to calculate the video update cycle; Furthermore, the specific steps of calculating the video update period by counting the time differences of all records are as follows: S301, according to the user habit list record history, the user watches a video of a type not recorded in the user habit list at each habit time, and determines the event as abnormal user emotion, and extracts the time of each event occurrence as , represents the time of the 1st, 2nd, 3rd, ..., jth event extracted, where j is a positive integer; calculate the time difference between two adjacent events, and design all the times to be calculated as , represents the calculated 1st, 2nd, 3rd, ..., j-1th adjacent time differences; S302: Draw a curve graph using the calculated adjacent time differences and corresponding time periods, use the adjacent time differences on the ordinate of the curve graph as the video update cycle, and construct dynamically changing video update cycles in different time periods.

[0009] S400, collecting the video switching frequency of the user when watching videos using the audio-visual device in the history, analyzing the video switching frequency of the user in the collected history, and calculating the normal threshold of the user's emotion; Furthermore, the specific steps for calculating the normal threshold of user emotions are as follows: S401, collecting the switching frequency of videos when users use audio and video equipment to watch videos in the history , Indicates the switching frequency of the video when the user uses the audio-visual device to watch the video for the 1st, 2nd, 3rd, ..., gth time in the collected history, where g is a positive integer; a curve graph is drawn using all the collected switching frequencies to generate a switching frequency curve; S402, performing a derivative operation on the generated switching frequency curve to generate a derivative curve, searching for a maximum peak value in the derivative curve, mapping the maximum peak value in the found derivative curve to the switching frequency curve, extracting a switching frequency Qy corresponding to the mapping point in the switching frequency curve, and using Qy as a normal threshold value of user emotion.

[0010] Analyze the switching frequency of videos when users watch videos in history, and use the switching frequency of users to reflect the user's emotional state. When the switching frequency is too high, it proves that the user is not satisfied with the played video and is eager to find the video he likes. The switching frequency can be used to calculate the user's normal emotional threshold; S500, when the user uses the audio-visual device to watch a video, the time when the user uses the audio-visual device is collected in real time, the marked time is checked by the real time in the user's habit list, and the video in the user's favorite group is automatically played; Furthermore, the specific steps of automatically playing videos in the user's favorite group are: S501. When a user uses an audio-visual device to watch a video, the time the user uses the audio-visual device is collected in real time as Ts, and the marked time is traversed and checked by real-time time in the user habit list. When Ts∈time, the video type corresponding to time is extracted and automatically played.

[0011] S600. Before automatically playing the videos in the user's favorite group, calculate the time difference between the real time and the last time the user's emotions were abnormal, determine whether the time difference has reached the video update cycle, and automatically play the videos in the pending group when it is determined to have reached it; if it has not reached it, use the user's emotions to determine the normal threshold, and then play the video again.

[0012] After judging the user's viewing time using the user habit list, provide the user's favorite video types; use personalized video playback mode to be more in line with the user's viewing habits and facilitate users to find their favorite videos. Attract users to use the platform for a long time, increase viewing time, and thus enhance user loyalty. Reduce the user's decision-making burden when choosing content, making the viewing experience smoother.

[0013] Furthermore, the specific steps for determining whether the time difference reaches the video update period are: S601, before automatically playing the video in the user's favorite group, extract the time of the last abnormal user emotion as Tj, calculate the time difference between the real time and the last abnormal user emotion as Tc, substitute the real time into the video update cycle curve, calculate the video update cycle corresponding to the real time as ST, judge the calculated time difference in the video update cycle, when When the video type needs to be updated, the video in the pending group is played; By judging the real-time time when users watch videos, when the video update cycle is reached, new videos in the pending group are played to users, increasing the diversity and novelty of videos watched by users; avoiding video fatigue and boredom caused by users watching the same type of videos all the time, and improving the user's viewing experience; S602, when When the real-time time is judged to have not reached the video update cycle, the videos in the hobby group continue to be automatically played; then the user's real-time emotions are judged, the real-time switching times and viewing time of the videos in the hobby group that the user watches automatically when using audio-visual equipment are extracted, and the real-time switching times are calculated. The formula is: ,In the formula, Qs represents the real-time switching frequency of the user to the video, K represents the number of real-time switching, and Gt represents the real-time viewing time of the user to the video in the hobby group; when When the user's real-time mood is abnormal, the automatic playback of videos in the favorite group is stopped, and the videos in the pending group are updated; when When the user's real-time mood is judged to be normal, the video in the hobby group continues to be played.

[0014] By judging the user's emotions in real time, when the user's emotions are abnormal, the playback of the favorite group videos will be stopped and the new videos in the pending group will be played instead. This will provide users with novel video types to prevent users from getting worse due to abnormal emotions, help users better immerse themselves in the viewing experience, enhance the viewing pleasure, improve the depth and accuracy of personalized recommendations, and create a more considerate user experience.

[0015] An intelligent management system for audio-visual equipment based on multi-source signals, which includes a data collection module, a video classification module, a user habit analysis module, a video update module and a user emotion analysis module; The data collection module is used to collect the switching frequency and viewing data of the user when watching videos using the audio and video device in the history; and collect all the videos played by the audio and video device in the history; The video classification module is used to analyze all videos played by audio and video devices in the collected history, build a deep learning model to classify all videos, build a video database, use the video database as a training set, input it into a support vector machine to train a video classifier, and classify all videos extracted from the network by audio and video devices in real time; The user habit analysis module is used to analyze the viewing data of all videos in the user's history, and to secondary classify the videos in the video database into a favorite group, a pending group, and a blocked group; and to mark the viewing time of the videos in the favorite group to generate a user habit list; The video update module is used to determine the real-time time when the user uses the audio-visual device to watch the video, and use the video update cycle to play the video in the pending group when the user watches the video to complete the video update; and update the video by judging the user's emotions; The user emotion analysis module is used to analyze the switching frequency of videos when the user uses audio-visual equipment to watch videos in the history, draw a curve graph, obtain the derivative of the curve, find the maximum peak of the derivative curve and map it in the switching frequency curve, use the mapping point to generate a normal threshold value of user emotions, and use the normal threshold value of emotions to judge the user's real-time emotions.

[0016] The video classification module includes a video database unit and a classifier construction unit; The video database unit is used to analyze all videos played by audio and video devices in the collected history, build a deep learning model to classify all videos, and build a video database; The classifier construction unit is used to input the video database as a training set into a support vector machine to train a video classifier, and to classify all videos extracted from the network by audio and video equipment in real time.

[0017] The video update module includes a periodic update unit and an abnormal emotion update unit; The periodic updating unit is used to determine the time when the user uses the audio-visual device in real time by calculating the video update period, and play the video in the pending group for updating; The abnormal emotion updating unit is used to judge the real-time emotion of the user by using the normal emotion threshold of the user, and when the user's emotion is abnormal, the video in the pending group is played to complete the video update.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses a video database to train a support vector machine to construct a video classifier, and uses the classifier to perform real-time classification after the audio-visual device extracts the video in the network; after the video is classified, the user can quickly find the video of interest according to different classifications when watching the video.

[0019] 2. The present invention provides the types of videos that users like; uses personalized video playback modes to better suit users' video viewing habits and facilitate users to find their favorite videos. It attracts users to use the platform for a long time, increases viewing time, and thus enhances user loyalty. It reduces the decision-making burden of users when selecting content, making the viewing experience smoother.

[0020] 3. The present invention determines the real-time time when the user watches the video, and when the video update cycle is reached, plays the new video in the pending group to the user, thereby increasing the diversity and novelty of the videos watched by the user; avoiding video fatigue and boredom caused by the user watching the same type of videos all the time, and improving the user's viewing experience.

[0021] 4. The present invention makes real-time judgment on user emotions. When the user's emotions are abnormal, the video in the favorite group is stopped from playing and the new video in the pending group is played instead. This provides users with novel video types, avoids users' aggravation due to abnormal emotions, helps users better immerse themselves in the viewing experience, enhances the viewing pleasure, and improves the depth and accuracy of personalized recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a module distribution diagram of the intelligent management system of audio-visual equipment based on multi-source signals of the present invention; Figure 2 It is a schematic diagram of the steps of the intelligent management method of audio-visual equipment based on multi-source signals of the present invention; Figure 3 It is an update cycle curve diagram of the intelligent management method of audio-visual equipment based on multi-source signals of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Example: Figure 1-Figure 3 As shown, the present invention provides a technical solution. An intelligent management method for audio-visual equipment based on multi-source signals, the method comprising the following steps: S100, collecting all videos played by audio and video devices in history, classifying all the collected videos in history using a deep learning model to generate a video database, and training the classified videos using a support vector machine to construct a video classifier; The specific steps of using support vector machine to train the classified videos and build a video classifier are as follows: S101. Collect all videos played by audio and video devices in history, and use convolutional neural network to extract features of each frame of each collected video. The formula is: ; In the formula, Represents the features of each frame of each video extracted, CNN represents the convolutional neural network algorithm, Represents each frame of each video; The recurrent neural network algorithm is used to combine the extracted features of each frame with the time sequence to perform time series modeling, and the time sequence of each frame in each video in the history is extracted. The formula is: ;In the formula, represents the constructed time series model, LSTM represents the recurrent neural network, Represents the characteristics of each frame, Indicates the time sequence corresponding to each frame; The deep learning model is constructed by combining the feature extraction and timing model of each frame. The generated deep learning model is In the deep learning model, out represents the output result, softmax represents the activation function, and Dense represents the fully connected layer. The constructed deep learning model is used to classify all the collected videos, and the classification results are , represents the first, second, third, ..., nth video types obtained by classification, where n is a positive integer; a video database is generated using the classified video categories; S102. After classifying the videos in history, the classified video database is input into the support vector machine as a training set to train a video classifier; the audio-visual equipment uses the wireless network to advance the video in the cloud, and after classifying the extracted video using the video classifier, it is transmitted to the video category corresponding to the video database.

[0025] Through deep learning models, all data played in audio-visual devices in history are classified to generate a video database, and the video database is used to train the support vector machine to build a video classifier. After the audio-visual device extracts the video from the network, the classifier is used to perform real-time classification. After the video is classified, users can quickly find videos of interest according to different classifications when watching videos.

[0026] S200, collecting viewing data of all videos of audio-visual device users in history, analyzing viewing data of all videos in user history, dividing all video categories in the video database into secondary categories, respectively constructing a favorite group, a blocked group, and a pending group; marking viewing time of users of video categories in the favorite group, and generating a user habit list; The specific steps to generate a user habit list are: S201, collect the viewing data of all videos by audio-visual device users in history, and extract the viewing times of each type of video in the viewing data. , Indicates the number of times the user has viewed the 1st, 2nd, 3rd, ..., nth type of video in history; after sorting the number of views from small to large, calculate the difference in the number of views of adjacent types of videos, and the difference in all the number of views is calculated. , It means that the differences of the 1st, 2nd, 3rd, ... n-1th adjacent viewing times are calculated, the n-1 viewing times differences are judged, the two difference generation nodes with the largest viewing times differences are selected, the first largest difference in the sorting is used as the node 1 of the shielded group and the pending group, and the second largest difference in the sorting is used as the node 2 of the shielded group and the favorite group; all video types before node 1 in the sorting are used as the shielded group; the video types between node 1 and node 2 in the sorting are used as the pending group; the video types after node 2 in the sorting are used as the favorite group; S201, extracting the viewing time of the video types in the generated hobby group in the history of the user, and marking each video type in the hobby group with the viewing time , Indicates the 1st, 2nd, 3rd, ..., mth video types in the favorite group after marking; generates a user habit list using the marked video types and time.

[0027] By using the viewing data of all types of videos in the user's history, we can analyze the user's preference for each type of video, and divide all videos into a favorite group, a pending group and a blocked group. The blocked group represents the types of videos that the user does not like, and the audio-visual equipment will block such videos and not play them; and the time the user watches videos in the favorite group is recorded to generate a user habit list; when the user uses the audio-visual equipment, personalized and humanized video recommendations are provided based on the user habit list, which greatly improves the user's comfort in watching videos.

[0028] S300, according to the user habit list record history, the user watches a video of a type not recorded in the user habit list at each habit time, the event is judged as abnormal user emotion, and the time difference of all records is counted to calculate the video update cycle; The specific steps of calculating the video update period by counting the time difference of all records are as follows: S301, according to the user habit list record history, the user watches a video of a type not recorded in the user habit list at each habit time, and determines the event as abnormal user emotion, and extracts the time of each event occurrence as , represents the time of the 1st, 2nd, 3rd, ..., jth event extracted, where j is a positive integer; calculate the time difference between two adjacent events, and design all the times to be calculated as , represents the calculated 1st, 2nd, 3rd, ..., j-1th adjacent time differences; S302: Draw a curve graph using the calculated adjacent time differences and corresponding time periods, use the adjacent time differences on the ordinate of the curve graph as the video update cycle, and construct dynamically changing video update cycles in different time periods.

[0029] S400, collecting the video switching frequency of the user when watching videos using the audio-visual device in the history, analyzing the video switching frequency of the user in the collected history, and calculating the normal threshold of the user's emotion; The specific steps for calculating the normal threshold of user emotions are: S401, collecting the switching frequency of videos when users use audio and video equipment to watch videos in the history , Indicates the switching frequency of the video when the user uses the audio-visual device to watch the video for the 1st, 2nd, 3rd, ..., gth time in the collected history, where g is a positive integer; a curve graph is drawn using all the collected switching frequencies to generate a switching frequency curve; S402, performing a derivative operation on the generated switching frequency curve to generate a derivative curve, searching for a maximum peak value in the derivative curve, mapping the maximum peak value in the found derivative curve to the switching frequency curve, extracting a switching frequency Qy corresponding to the mapping point in the switching frequency curve, and using Qy as a normal threshold value of user emotion.

[0030] Analyze the switching frequency of videos when users watch videos in history, and use the switching frequency of users to reflect the user's emotional state. When the switching frequency is too high, it proves that the user is not satisfied with the played video and is eager to find the video he likes. The switching frequency can be used to calculate the user's normal emotional threshold; S500, when the user uses the audio-visual device to watch a video, the time when the user uses the audio-visual device is collected in real time, the marked time is checked by the real time in the user's habit list, and the video in the user's favorite group is automatically played; The specific steps to automatically play videos in a user's favorite group are: S501. When a user uses an audio-visual device to watch a video, the time the user uses the audio-visual device is collected in real time as Ts, and the marked time is traversed and checked by real-time time in the user habit list. When Ts∈time, the video type corresponding to time is extracted and automatically played.

[0031] S600. Before automatically playing the videos in the user's favorite group, calculate the time difference between the real time and the last time the user's emotions were abnormal, determine whether the time difference has reached the video update cycle, and automatically play the videos in the pending group when it is determined to have reached it; if it has not reached it, use the user's emotions to determine the normal threshold, and then play the video again.

[0032] After judging the user's viewing time using the user habit list, provide the user's favorite video types; use personalized video playback mode to be more in line with the user's viewing habits and facilitate users to find their favorite videos. Attract users to use the platform for a long time, increase viewing time, and thus enhance user loyalty. Reduce the user's decision-making burden when choosing content, making the viewing experience smoother.

[0033] The specific steps for determining whether the time difference reaches the video update period are: S601, before automatically playing the video in the user's favorite group, extract the time of the last abnormal user emotion as Tj, calculate the time difference between the real time and the last abnormal user emotion as Tc, substitute the real time into the video update cycle curve, calculate the video update cycle corresponding to the real time as ST, judge the calculated time difference in the video update cycle, when When the video type needs to be updated, the video in the pending group is played; By judging the real-time time when users watch videos, when the video update cycle is reached, new videos in the pending group are played to users, increasing the diversity and novelty of videos watched by users; avoiding video fatigue and boredom caused by users watching the same type of videos all the time, and improving the user's viewing experience; S602, when When the real-time time is judged to have not reached the video update cycle, the videos in the hobby group continue to be automatically played; then the user's real-time emotions are judged, the real-time switching times and viewing time of the videos in the hobby group that the user watches automatically when using audio-visual equipment are extracted, and the real-time switching times are calculated. The formula is: ,In the formula, Qs represents the real-time switching frequency of the user to the video, K represents the number of real-time switching, and Gt represents the real-time viewing time of the user to the video in the hobby group; when When the user's real-time mood is abnormal, the automatic playback of videos in the favorite group is stopped, and the videos in the pending group are updated; when When the user's real-time mood is judged to be normal, the video in the hobby group continues to be played.

[0034] By judging the user's emotions in real time, when the user's emotions are abnormal, the playback of the favorite group videos will be stopped and the new videos in the pending group will be played instead. This will provide users with novel video types to prevent users from getting worse due to abnormal emotions, help users better immerse themselves in the viewing experience, enhance the viewing pleasure, improve the depth and accuracy of personalized recommendations, and create a more considerate user experience.

[0035] An intelligent management system for audio-visual equipment based on multi-source signals, which includes a data collection module, a video classification module, a user habit analysis module, a video update module and a user emotion analysis module; The data collection module is used to collect the switching frequency and viewing data of the user when watching videos using the audio and video device in the history; and collect all the videos played by the audio and video device in the history; The video classification module is used to analyze all videos played by audio and video devices in the collected history, build a deep learning model to classify all videos, build a video database, use the video database as a training set, input it into a support vector machine to train a video classifier, and classify all videos extracted from the network by audio and video devices in real time; The user habit analysis module is used to analyze the viewing data of all videos in the user's history, and to secondary classify the videos in the video database into a favorite group, a pending group, and a blocked group; and to mark the viewing time of the videos in the favorite group to generate a user habit list; The video update module is used to determine the real-time time when the user uses the audio-visual device to watch the video, and use the video update cycle to play the video in the pending group when the user watches the video to complete the video update; and update the video by judging the user's emotions; The user emotion analysis module is used to analyze the switching frequency of videos when the user uses audio-visual equipment to watch videos in the history, draw a curve graph, obtain the derivative of the curve, find the maximum peak of the derivative curve and map it in the switching frequency curve, use the mapping point to generate a normal threshold value of user emotions, and use the normal threshold value of emotions to judge the user's real-time emotions.

[0036] The video classification module includes a video database unit and a classifier construction unit; The video database unit is used to analyze all videos played by audio and video devices in the collected history, build a deep learning model to classify all videos, and build a video database; The classifier construction unit is used to input the video database as a training set into a support vector machine to train a video classifier, and to classify all videos extracted from the network by audio and video equipment in real time.

[0037] The video update module includes a periodic update unit and an abnormal emotion update unit; The periodic updating unit is used to determine the time when the user uses the audio-visual device in real time by calculating the video update period, and play the video in the pending group for updating; The abnormal emotion updating unit is used to judge the real-time emotion of the user by using the normal emotion threshold of the user, and when the user's emotion is abnormal, the video in the pending group is played to complete the video update.

[0038] Embodiment: Analyze the audio-visual equipment in the home and classify the videos played by the audio-visual equipment into science and education, art, entertainment, and animation categories; classify the videos on the Internet by building a classifier; Analyze the viewing data of users in history and classify all videos into two categories: the hobby group is entertainment and animation, the pending group is art, and the blocked group is science and education. Mark the video viewing time in the user's hobby group, and get entertainment: 5:00pm-7:00pm, animation: 7:00pm-9:00pm to build a user habit list. The video update cycle is calculated based on the historical data of users watching videos. Figure 3 As shown, the last time the user's emotions were abnormal was on April 10; On April 15, the user went home to watch videos at 6 pm. The corresponding video update cycle is 9 days in the video update cycle curve chart, and the real-time time difference is 5 days, so there is no need to update the video. The corresponding time is found in the user habit list as entertainment, and entertainment videos are automatically played; Assume that the normal threshold of user emotions calculated based on historical data is 4 times / min, and the real-time user video switching frequency is 10 times / min; if the user's emotions are abnormal, stop playing entertainment videos and update to play art videos.

[0039] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An intelligent management method for audio-visual equipment based on multi-source signals, characterized in that: The method comprises the following steps: S100, collecting all videos played by audio and video devices in history, classifying all the collected videos in history using a deep learning model to generate a video database, and training the classified videos using a support vector machine to construct a video classifier; S200, collecting viewing data of all videos of audio-visual device users in history, analyzing viewing data of all videos in user history, dividing all video categories in the video database into secondary categories, respectively constructing a favorite group, a blocked group, and a pending group; marking viewing time of users of video categories in the favorite group, and generating a user habit list; S300, according to the user habit list record history, the user watches a video of a type not recorded in the user habit list at each habit time, the event is judged as abnormal user emotion, and the time difference of all records is counted to calculate the video update cycle; S400, collecting the video switching frequency of the user when watching videos using the audio-visual device in the history, analyzing the video switching frequency of the user in the collected history, and calculating the normal threshold of the user's emotion; S500, when the user uses the audio-visual device to watch a video, the time when the user uses the audio-visual device is collected in real time, the marked time is checked by the real time in the user's habit list, and the video in the user's favorite group is automatically played; S600. Before automatically playing the videos in the user's favorite group, calculate the time difference between the real time and the last time the user's emotions were abnormal, determine whether the time difference has reached the video update cycle, and automatically play the videos in the pending group when it is determined to have reached it; if it has not reached it, use the user's emotions to determine the normal threshold, and then play the video again.

2. The intelligent management method of audio-visual equipment based on multi-source signals according to claim 1, characterized in that: The specific steps of using the support vector machine to train the classified video and construct a video classifier in S100 are: S101. Collect all videos played by audio and video devices in history, and use convolutional neural network to extract features of each frame of each collected video. The formula is: ; In the formula, Represents the features of each frame of each video extracted, CNN represents the convolutional neural network algorithm, Represents each frame of each video; The recurrent neural network algorithm is used to combine the extracted features of each frame with the time sequence to perform time series modeling, and the time sequence of each frame in each video in the history is extracted. The formula is: ;In the formula, represents the constructed time series model, LSTM represents the recurrent neural network, Represents the characteristics of each frame, Indicates the time sequence corresponding to each frame; The deep learning model is constructed by combining the feature extraction and timing model of each frame. The generated deep learning model is In the deep learning model, out represents the output result, softmax represents the activation function, and Dense represents the fully connected layer. The constructed deep learning model is used to classify all the collected videos, and the classification results are , represents the first, second, third, ..., nth video types obtained by classification, where n is a positive integer; a video database is generated using the classified video categories; S102. After classifying the videos in history, the classified video database is input into the support vector machine as a training set to train a video classifier; the audio-visual equipment uses the wireless network to advance the video in the cloud, and after classifying the extracted video using the video classifier, it is transmitted to the video category corresponding to the video database.

3. The intelligent management method of audio-visual equipment based on multi-source signals according to claim 1, characterized in that: The specific steps of generating the user habit list in S200 are: S201, collect the viewing data of all videos by audio-visual device users in history, and extract the viewing times of each type of video in the viewing data. , Indicates the number of times the user has viewed the 1st, 2nd, 3rd, ..., nth type of video in history; after sorting the number of views from small to large, calculate the difference in the number of views of adjacent types of videos, and the difference in all the number of views is calculated. , It means that the differences of the 1st, 2nd, 3rd, ... n-1th adjacent viewing times are calculated, the n-1 viewing times differences are judged, the two difference generation nodes with the largest viewing times differences are selected, the first largest difference in the sorting is used as the node 1 of the shielding group and the pending group, and the second largest difference in the sorting is used as the node 2 of the shielding group and the hobby group; all video types before node 1 in the sorting are used as the shielding group; The video types between node 1 and node 2 in the sorting are taken as the pending group; The video types after node 2 in the sorting are regarded as the hobby group; S201, extracting the viewing time of the video types in the generated hobby group in the history of the user, and marking each video type in the hobby group with the viewing time , Indicates the 1st, 2nd, 3rd, ..., mth video types in the favorite group after marking; generates a user habit list using the marked video types and time.

4. The intelligent management method of audio-visual equipment based on multi-source signals according to claim 1, characterized in that: The specific steps of calculating the video update period by counting the time differences of all records in S300 are as follows: S301, according to the user habit list record history, the user watches a video of a type not recorded in the user habit list at each habit time, and determines the event as abnormal user emotion, and extracts the time of each event occurrence as , represents the time of the 1st, 2nd, 3rd, ..., jth event extracted, where j is a positive integer; calculate the time difference between two adjacent events, and design all the times to be calculated as , represents the calculated 1st, 2nd, 3rd, ..., j-1th adjacent time differences; S302: Draw a curve graph using the calculated adjacent time differences and corresponding time periods, use the adjacent time differences on the ordinate of the curve graph as the video update cycle, and construct dynamically changing video update cycles in different time periods.

5. The intelligent management method of audio-visual equipment based on multi-source signals according to claim 1, characterized in that: The specific steps of calculating the normal threshold of the user's emotion in S400 are: S401, collecting the switching frequency of videos when users use audio and video equipment to watch videos in the history , Indicates the switching frequency of the video when the user uses the audio-visual device to watch the video for the 1st, 2nd, 3rd, ..., gth time in the collected history, where g is a positive integer; a curve graph is drawn using all the collected switching frequencies to generate a switching frequency curve; S402, performing a derivative operation on the generated switching frequency curve to generate a derivative curve, searching for a maximum peak value in the derivative curve, mapping the maximum peak value in the found derivative curve to the switching frequency curve, extracting a switching frequency Qy corresponding to the mapping point in the switching frequency curve, and using Qy as a normal threshold value of user emotion.

6. The intelligent management method of audio-visual equipment based on multi-source signals according to claim 3, characterized in that: The specific steps of automatically playing the videos in the user's favorite group in S500 are: S501. When a user uses an audio-visual device to watch a video, the time the user uses the audio-visual device is collected in real time as Ts, and the marked time is traversed and checked by real-time time in the user habit list. When Ts∈time, the video type corresponding to time is extracted and automatically played.

7. The intelligent management method of audio-visual equipment based on multi-source signals according to claim 5, characterized in that: The specific steps of determining whether the time difference reaches the video update period in S600 are: S601, before automatically playing the video in the user's favorite group, extract the time of the last abnormal user emotion as Tj, calculate the time difference between the real time and the last abnormal user emotion as Tc, substitute the real time into the video update cycle curve, calculate the video update cycle corresponding to the real time as ST, judge the calculated time difference in the video update cycle, when When the video type needs to be updated, the video in the pending group is played; S602, when When the real-time time is judged to have not reached the video update cycle, the videos in the hobby group continue to be automatically played; then the user's real-time emotions are judged, the real-time switching times and viewing time of the videos in the hobby group that the user watches automatically when using audio-visual equipment are extracted, and the real-time switching times are calculated. The formula is: ,In the formula, Qs represents the real-time switching frequency of the user to the video, K represents the number of real-time switching, and Gt represents the real-time viewing time of the user to the video in the hobby group; when When the user's real-time mood is abnormal, the automatic playback of videos in the favorite group is stopped, and the videos in the pending group are updated; when When the user's real-time mood is judged to be normal, the video in the hobby group continues to be played.

8. Intelligent management system for audio-visual equipment based on multi-source signals, characterized by: The intelligent management system for audio-visual equipment includes a data collection module, a video classification module, a user habit analysis module, a video update module, and a user emotion analysis module; The data collection module is used to collect the switching frequency and viewing data of the user when watching videos using the audio and video device in the history; and collect all the videos played by the audio and video device in the history; The video classification module is used to analyze all videos played by audio and video devices in the collected history, build a deep learning model to classify all videos, build a video database, use the video database as a training set, input it into a support vector machine to train a video classifier, and classify all videos extracted from the network by audio and video devices in real time; The user habit analysis module is used to analyze the viewing data of all videos in the user's history, and to secondary classify the videos in the video database into a favorite group, a pending group, and a blocked group; and to mark the viewing time of the videos in the favorite group to generate a user habit list; The video update module is used to determine the real-time time when the user uses the audio-visual device to watch the video, and use the video update cycle to play the video in the pending group when the user watches the video to complete the video update; and update the video by judging the user's emotions; The user emotion analysis module is used to analyze the switching frequency of videos when the user uses audio-visual equipment to watch videos in the history, draw a curve graph, obtain the derivative of the curve, find the maximum peak of the derivative curve and map it in the switching frequency curve, use the mapping point to generate a normal threshold value of user emotions, and use the normal threshold value of emotions to judge the user's real-time emotions.

9. The intelligent management system for audio-visual equipment based on multi-source signals according to claim 8, characterized in that: The video classification module includes a video database unit and a classifier construction unit; The video database unit is used to analyze all videos played by audio and video devices in the collected history, build a deep learning model to classify all videos, and build a video database; The classifier construction unit is used to input the video database as a training set into a support vector machine to train a video classifier, and to classify all videos extracted from the network by audio and video equipment in real time.

10. The intelligent management system for audio-visual equipment based on multi-source signals according to claim 8, characterized in that: The video update module includes a periodic update unit and an abnormal emotion update unit; The periodic updating unit is used to determine the time when the user uses the audio-visual device in real time by calculating the video update period, and play the video in the pending group for updating; The abnormal emotion updating unit is used to judge the real-time emotion of the user by using the normal emotion threshold of the user, and when the user's emotion is abnormal, the video in the pending group is played to complete the video update.