Personalized video recommendation system based on use habits of user

Abstract

The invention is suitable for the technical field of customized screening systems, and provides a personalized video recommendation system based on use habits of a user. The system comprises an environment and time identification module, a visitor emotion perception module, an audio-visual material screening algorithm module, a visitor feedback learning module, a visitor interface and interaction module and a visitor privacy protection module. And the screening content is adjusted in real time in combination with the environment characteristics and the situation states of the visitors, so that the screening is highly related to the current situations of the visitors. Through facial feature analysis and body language analysis, customization requirements of visitors are more accurately met, and the method can adapt to dynamic interests of the visitors changing due to various factors such as time, places, conditions and social events.

Description

Technical Field

[0001] The present invention belongs to the technical field of customized screening systems, and particularly relates to a personalized video recommendation system based on user usage habits. Background Art

[0002] In today's field of customized audio-visual material screening, the application of deep learning has become the industry standard. It uses advanced neural network models such as convolutional neural networks and recurrent neural networks to help the screening system more deeply analyze visitor preferences and audio-visual material content. In addition, screening engines are now also committed to multi-modal data fusion, which means that the system not only processes visitor interaction data, but also integrates the text descriptions, images, audio of audio-visual materials, and the audio-visual material features themselves to further improve the accuracy of screening. To keep up with the immediacy of visitor needs, real-time screening technology is also widely adopted to ensure that the screened content can be quickly updated as soon as the visitor's behavior changes.

[0003] Although the technology of customized audio-visual material screening systems is becoming increasingly mature, it does face diverse challenges, especially how to adapt to the constantly changing interests and needs of visitors. Traditionally, these systems rely on visitors' previous browsing data and search records for screening, and this method may confine visitors to past behaviors and make it difficult for them to access new content areas. In addition, visitors' interests are not static and will change due to various factors such as time, location, mood, and social events. Summary of the Invention

[0004] The purpose of the present invention is to provide a personalized video recommendation system based on user usage habits, aiming to solve the technical problems existing in the prior art identified in the background art.

[0005] The present invention is implemented as follows. A personalized video recommendation system based on user usage habits, the system includes:

[0006] An environment and time recognition module, used to obtain the location information of the visitor through the GPS of the portable multimedia player, read the system time, and collect ambient volume and light intensity data through the microphone and sensor of the portable multimedia player to assist in determining the environment where the visitor is currently located;

[0007] A visitor mood perception module, used to capture the dynamic facial features of the visitor through the front camera, capture the body language of the visitor, determine the visitor's state, and combine the data of the dynamic facial features and body language to create the mood temperature state of the visitor;

[0008] An audio-visual material screening algorithm module, which is used to construct a customized visitor profile by combining the visitor's previous review data and the current mood temperature state, and according to the visitor profile and the current mood temperature state, integrating location, time, and environmental factors into the screening logic to match the content that meets the screening logic from the audio-visual material library;

[0009] A visitor feedback learning module, which is used to monitor the visitor's review of the screened audio-visual materials, record the completeness of the visitor's review of the screened audio-visual materials, optimize the visitor profile and the screening algorithm, and provide a feedback mechanism for the visitor.

[0010] As a further solution of the present invention, the environment and time recognition module specifically includes:

[0011] A location positioning unit, which is used to obtain the visitor's current geographical coordinates using the GPS function of the portable multimedia player, and based on the GPS data, combined with the geographical information system, judge the macroscopic environment where the visitor is located, denoted as spatial location data;

[0012] A time acquisition unit, which is used to directly obtain the current time from the device's system clock and analyze which period of the day the visitor is in, denoted as time period data;

[0013] An environment perception unit, which is used to measure the noise level of the surrounding environment using the device's microphone, denoted as noise level data, and use the light sensor and the camera to evaluate the light intensity of the current environment, denoted as light intensity data;

[0014] A visitor environment determination unit, which integrates the spatial location data, time period data, noise level data, and light intensity data to determine the characteristics of the specific environment where the visitor is located. The specific formula is:

[0015] H = w1×S + w2×T + W3×N + W4×L

[0016] H is the environmental characteristic score, S is the spatial location data, T is the time period data, N is the noise level data, L is the light intensity data, and w1, w2, w2, w3 are weighting factors.

[0017] As a further solution of the present invention, the visitor mood perception module specifically includes:

[0018] A facial feature recognition unit, which is used to use the front camera of the device to capture the facial feature image of the visitor in real time, locate the facial features of the visitor, and analyze the facial feature data to identify the basic mood represented by the visitor's features, denoted as facial feature mood data;

[0019] A body language analysis unit, which is used to capture the body movements and postures of visitors, identify the key points of the body, analyze the positions and movements of the body key points, identify the body language of visitors, and record it as body language modality data;

[0020] A modality temperature calculation unit, which is used for facial feature modality data and body language modality data, calculates the modality intensity of visitors, and records it as a modality temperature score. The specific formula is:

[0021] Y = ɑ×E + β×B + γ

[0022] Y is the modality temperature score, ɑ and β are weight factors, E is the facial feature modality data, B is the body language modality data, and γ is the baseline offset;

[0023] A visitor modality pattern learning unit, which is used to collect and analyze the previous modality data of visitors, learn the modality patterns of visitors, and predict the possible modality changes of visitors.

[0024] As a further solution of the present invention, the audio-visual material screening algorithm module specifically includes:

[0025] A visitor profile construction unit, which is used to collect the previous review data of visitors, including the types of audio-visual materials reviewed, review time, review frequency, analyze the preferences and habits of visitors, and construct a customized profile of visitors;

[0026] A screening strategy determination unit, which is used to determine the content categories and types to be screened according to the visitor profile, environmental feature score, and modality temperature score. The specific formula is:

[0027] [R(U, V) = p1×H_user + p2×Y_user + p3×P(U, V) + p4×C(U, V) + δ] where R(U, V) is the screening strategy determination score, U is the visitor, V is the candidate audio-visual material, H_user is the feature score of the current environment of the visitor, from the environment and time recognition module, Y_user is the modality temperature of the visitor, from the visitor modality perception module, p1, p2, p3, p4 are weight parameters, δ is a constant bias term, P(U, V) is the preference similarity calculation score between the previous review data of visitor U and candidate audio-visual material V, and C(U, V) is the content analysis score, according to the matching degree between the content of the audio-visual material and the visitor's preferences;

[0028] A screening content matching unit, the goal of the algorithm is to maximize the R(U, V) score, that is, the expected satisfaction of visitor U with audio-visual material V, and screen a series of audio-visual materials with the highest scores for the visitor.

[0029] As a further solution of the present invention, the visitor feedback learning module specifically includes:

[0030] A feedback learning optimization unit, which is used to monitor the completeness of visitor review, review time, and the reaction of the visitor score to the screened content, analyze the visitor feedback, and dynamically adjust and optimize the visitor profile and screening strategy;

[0031] A visitor interaction and feedback unit, which is used to provide a visitor interface, display the screened audio-visual materials, collect direct feedback from visitors such as liking, collecting, and commenting on the screened content, and further optimize the visitor profile and screened content through the interaction behavior of visitors.

[0032] As a further solution of the present invention, it further includes a visitor interface and interaction module, which specifically includes:

[0033] An audio-visual material screening and display unit, which is used to display the list of screened audio-visual materials;

[0034] A visitor customization setting unit, which is used to provide personal information management, view and edit personal profiles, set preference adjustment functions, and specify and adjust the weights of factors in the screening algorithm.

[0035] As a further solution of the present invention, it further includes a visitor privacy protection module, which specifically includes:

[0036] A visitor consent and data collection unit, which is used to display the purpose and scope of data collection through the visitor interface and obtain collection authorization before using any data collection function;

[0037] A data security and encryption unit, which is used to encrypt all collected data.

[0038] The beneficial effects of the present invention are as follows: This system not only relies on the previous browsing behavior of visitors, but also combines the environmental characteristics and emotional states of visitors to adjust the screened content in real time, so as to ensure that the screening is highly relevant to the current situation of visitors. Through facial feature analysis and body language analysis, it can more accurately meet the customized needs of visitors and can adapt to the dynamic visitor interests that change due to various factors such as time, location, mood, and social events. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a structural block diagram of a personalized video recommendation system based on user usage habits provided by an embodiment of the present invention;

[0040] Figure 2 It is a structural block diagram of an environment and time recognition module provided by an embodiment of the present invention;

[0041] Figure 3The structural block diagram of the visitor mood perception module provided by the embodiment of the present invention;

[0042] Figure 4 The structural block diagram of the audio-visual material screening algorithm module provided by the embodiment of the present invention;

[0043] Figure 5 The structural block diagram of the visitor feedback learning module provided by the embodiment of the present invention;

[0044] Figure 6 The structural block diagram of the visitor interface and interaction module provided by the embodiment of the present invention;

[0045] Figure 7 The structural block diagram of the visitor privacy protection module provided by the embodiment of the present invention; Detailed implementation manners

[0046] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0047] It can be understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish the first element from another element. For example, without departing from the scope of this application, the first xx script may be referred to as the second xx script, and similarly, the second xx script may be referred to as the first xx script.

[0048] Figure 1 The structural block diagram of a personalized video recommendation system based on user usage habits provided by the embodiment of the present invention, as Figure 1 shown, a personalized video recommendation system based on user usage habits, the system includes:

[0049] An environment and time recognition module 100, configured to obtain the location information of a visitor through the GPS of a portable multimedia player, read the system time, and collect ambient volume and light intensity data through the microphone and sensor of the portable multimedia player to assist in determining the environment where the visitor is currently located;

[0050] This module uses the GPS function of a portable multimedia player to obtain the precise geographical location of visitors. It can be combined with a Geographic Information System (GIS) to not only determine the macro environment of visitors (such as cities, rural areas, commercial areas, residential areas, etc.), but also identify more detailed location information (such as parks, coffee shops, offices, etc.). The positioning technology can also include Wi-Fi positioning and cellular positioning to improve the accuracy of indoor positioning. The location data can be used to analyze the living patterns and behavioral habits of visitors, thereby inferring possible activity types and visitor preferences. The time is obtained from the system clock of the portable device and classified according to time periods, such as early morning, afternoon, evening, or late at night. The time data helps to infer the daily activities of visitors, such as working hours, rest times, or meal times, and filter the audiovisual material content based on this information.

[0051] This module measures the ambient noise level using the device's microphone, which can help determine whether the visitor is in a quiet home environment, a noisy outdoor environment, or a working environment. The light sensor and camera evaluate the light intensity of the current environment, such as bright outdoor light or dim indoor lighting. The environmental data helps the system infer the possible activity status of the visitor. For example, when outdoors, the visitor may be more inclined to review short audiovisual materials, while in a quiet indoor environment, the visitor may prefer to review long movies or TV series. The spatial location data, time period data, noise level data, and light intensity data are combined and a weighted formula is used to calculate an environmental characteristic score. The weight factors w1, w2, w2, w3 need to be dynamically adjusted based on data analysis and visitor feedback to ensure that the environmental score can accurately reflect the impact on the visitor's audiovisual material review habits. Through machine learning models and visitor feedback, the system can continuously self-optimize these weight factors to ensure that the filtered content best matches the visitor's current environment and mood state.

[0052] The visitor mood perception module 200 is used to capture the dynamic facial features of the visitor through the front camera, capture the visitor's body language, determine the visitor's state, and combine the data of the dynamic facial features and body language to create the mood temperature state of the visitor;

[0053] This module uses the front camera to capture the facial images of visitors in real time, and applies advanced image processing techniques to locate facial features such as eyes, eyebrows, mouth, and nose. Machine learning and deep learning algorithms, such as convolutional neural networks (CNNs), are applied to analyze the subtle changes in facial features to identify the micro-features of the visitors' moods. The facial features are mapped to basic mood categories such as happy, sad, angry, surprised, etc., and corresponding mood scores are assigned. The body postures and movements of visitors are captured by the camera, and pose estimation techniques are used to identify the key points of the body, such as hands, shoulders, head, and torso. The position changes of these key points over time are analyzed to identify limb movements such as crossed arms, rapid waving, etc., which may indicate specific mood states. Context information, such as whether the visitor's sitting posture is relaxed or standing is tense, is combined to further refine the identification of mood states.

[0054] This module designs evaluation functions E and B, which convert facial features and body language into quantified mood data respectively. Appropriate weight factors ɑ and β are determined, which may be obtained through experiments or data analysis, to ensure that they reflect the true proportions of facial and body language in mood expression. A baseline offset γ is set as the starting point of the mood thermometer, which helps to standardize the mood states of different time periods and different visitors. A time series analysis model is established to track the mood change process of visitors and learn the mood cycles and patterns of visitors. The visitor mood data is combined with the previous data of the visitor's review of audio-visual material content to discover the audio-visual material preferences of visitors in specific mood states. Through long-term observation and analysis, the mood change trends of visitors at specific times, in specific environments or situations are predicted. In addition to the explicit feedback from visitors (such as likes, comments, shares, etc.), feedback can also be implicitly obtained through the mood changes of visitors. The mood changes of visitors after reviewing different audio-visual material content are analyzed to adjust and optimize the screening algorithm.

[0055] The audio-visual material screening algorithm module 300 is used to combine the visitor's previous review data and the current mood temperature state to construct a customized visitor profile. According to the visitor profile and the current mood temperature state, and integrating location, time, and environmental factors into the screening logic, it matches the content that meets the screening logic from the audio-visual material library;

[0056] This module analyzes the previous review records of visitors through deep learning models such as recurrent neural networks (RNNs) or long short-term memory networks (LSTMs), and extracts patterns of review habits and content preferences in the time series. Natural language processing (NLP) techniques can be introduced to analyze the comments and feedback of visitors on the audio-visual material platform, further enriching the dimensions of the visitor profile. The dynamic update mechanism of the visitor profile enables the profile to reflect the latest preferences and behavior changes of visitors in real time. The multi-armed bandit algorithm or reinforcement learning is used to optimize the weight parameters p1, p2, p3, p4 to achieve self-learning and adjustment of the screening system.

[0057] Considering that the preferences of visitors are diverse and dynamically changing, the screening strategy adjusts the weight parameters regularly to adapt to the changes in visitor behavior. A context-aware screening mechanism is introduced to adjust the screening strategy according to different factors such as time, location, ambient volume, and light intensity, providing scenario-based screening of audio-visual materials. The collaborative filtering and content-based recommendation algorithms are used to comprehensively consider the matching degree between visitors and the content of audio-visual materials. Machine learning techniques such as matrix factorization are used to perform feature embedding on visitors and the content of audio-visual materials, so as to accurately calculate the similarity between visitor preferences and the content of audio-visual materials. The dynamic update mechanism is implemented so that the calculation of preference similarity can reflect the latest visitor behavior and content updates. Convolutional neural networks (CNNs) and natural language processing (NLP) techniques in deep learning are used to analyze the visual and text information of the content of audio-visual materials. Semantic analysis is used to understand the theme, plot, and emotional tendency of the audio-visual material, and these analysis results are compared with the visitor's emotional temperature and profile to evaluate the customized relevance of the content. Based on the above screening logic, the learning to rank method is applied to train the model so that it can output a sorted list of audio-visual materials with the highest expected satisfaction. A feedback loop mechanism is established to evaluate the quality of the screening results based on the behavior data of visitors such as review time, frequency, and interaction, and the screening algorithm is adjusted accordingly. A real-time feedback mechanism is introduced. For example, the behavior of visitors skipping the screened audio-visual materials can immediately adjust the screening list of the currently playing audio-visual materials. A / B testing and multivariate testing are used to continuously optimize the screening strategy to ensure the agility and adaptability of the screening system.

[0058] The visitor feedback learning module 400 is used to monitor visitors' review and screening of audio-visual materials, record the completeness of visitors' review of the screened audio-visual materials, optimize the visitor profile and screening algorithm, and provide a feedback mechanism for visitors.

[0059] This module uses deep learning technology to analyze the behavioral patterns of visitors when reviewing audio-visual materials, such as review time points, pauses, skips, replays, review progress, etc. Using natural language processing technology, sentiment analysis is performed on the comments and ratings of visitors to understand the positive or negative sentiment of visitors towards the content of the audio-visual materials. By analyzing this behavioral and linguistic data, this unit can convert the implicit feedback of visitors (such as review behavior) and explicit feedback (such as ratings and comments) into specific improvement measures for visitor profiles and screening algorithms. Real-time adjustment of the screened content is achieved. For example, if visitors continuously skip a certain type of audio-visual material, the system will immediately reduce the screening weight of this type of audio-visual material. Combining real-time data stream processing technology can quickly respond to the behavioral changes of visitors and adjust the currently playing screening list without affecting the visitor experience. Interactive questionnaires or gamification elements are provided to encourage visitors to provide more preference information. For example, visitors can choose their favorite types or themes of audio-visual materials through simple interactive games, helping the system to more accurately build visitor profiles. Using the choices of visitors in the questionnaire or game, refine the parameter adjustment of the screening algorithm to increase the degree of customization of the screening. Over time, analyze the long-term trends of visitor behavior, determine the patterns of changes in visitor preferences, and predict their future behavior. Combining time series analysis and the evolution model of visitor profiles, adjust the screening algorithm to ensure that it can reflect the development trend of visitor preferences.

[0060] This module integrates the activity information of visitors on social media, such as the preferences and sharing behaviors of visitors on other platforms, to provide a cross-platform customized experience. This technology can expand the sources of visitor feedback and further improve visitor profiles through external social signals. By continuously testing different screening strategies, the system can learn which strategies are most effective for different types of visitors and adjust its methods accordingly. Apply multi-armed bandit algorithms, reinforcement learning, etc., to continuously self-optimize to achieve the best screening effect. Evaluate visitor engagement, such as the number of reviews, review duration, etc., and visitor satisfaction, such as the repeat review rate, acceptance degree of the screened content, etc. Combine the engagement and satisfaction data of visitors to comprehensively score the screening system and guide future screening strategy adjustments.

[0061] Figure 2 For the structural block diagram of the environment and time recognition module provided by the embodiment of the present invention, as Figure 2 shown, the environment and time recognition module specifically includes:

[0062] A location positioning unit 110, which is used to obtain the current geographical coordinates of the visitor using the GPS function of the portable multimedia player, and based on the GPS data, combined with the geographic information system, judge the macro environment where the visitor is located, denoted as spatial location data;

[0063] A time acquisition unit 120, which is used to directly obtain the current time from the system clock of the device and analyze which period of the day the visitor is in, denoted as period data;

[0064] An environment perception unit 130, which is used to measure the noise level of the surrounding environment by using the microphone of the device, denoted as noise level data, and use a light sensor and a camera to evaluate the light intensity of the current environment, denoted as light intensity data;

[0065] A visitor environment determination unit 140, which integrates the spatial location data, period data, noise level data, and light intensity data to determine the characteristics of the specific environment where the visitor is located. The specific formula is:

[0066] H = W1×S + W2×T + W3×N + W4×L

[0067] S (spatial location data): A function for scoring the current macro environment based on geographical coordinates and a geographical information system, such as outdoors, commercial area, residential area, etc.

[0068] T (period data): Give a specific score to the environment according to the current period (such as morning, afternoon, evening).

[0069] N (noise level data): The score given according to the noise level, and there are significant differences in the noise levels of different environments.

[0070] L (light intensity data): The score given according to the light intensity, which can reflect the indoor / outdoor or lighting conditions at a specific time.

[0071] w1, w2, w2, w3: These are weight factors, which are determined according to the different importance of different data types for environment determination and need to be adjusted through data analysis and experiments to optimize the model performance.

[0072] Figure 3 It is a structural block diagram of the visitor mood perception module provided by the embodiment of the present invention. As Figure 3 shown, the visitor mood perception module specifically includes:

[0073] A facial feature recognition unit 210, which is used to use the front camera of the device to capture the facial feature image of the visitor in real time, locate the facial features of the visitor, analyze the facial feature data, and identify the basic mood represented by the visitor's features, denoted as facial feature mood data;

[0074] A body language analysis unit 220, which is used to capture the body movements and postures of the visitor, identify the key points of the body, analyze the positions and movements of the body key points, and identify the body language of the visitor, denoted as body language mood data;

[0075] A Mood Temperature Calculation Unit 230, which is used for facial feature mood data and body language mood data to calculate the mood intensity of the visitor, denoted as the mood temperature score. The specific formula is as follows:

[0076] Y = ɑ × E + β × B + γ

[0077] Y is the mood temperature score;

[0078] E (facial feature mood data): It is an evaluation function that converts the captured facial features into mood data, which may include the recognition and scoring of basic moods such as happiness, sadness, anger, surprise, etc.

[0079] B (body language mood data): It is another evaluation function that scores according to the visitor's body language, such as posture, gestures, etc., which can provide non-verbal cues of the mood state.

[0080] α and β: These are weight factors that determine the relative importance of facial features and body language in the final mood temperature. These factors are usually adjusted based on previous research or through data analysis.

[0081] γ: This is a baseline offset, which can be understood as the starting point for measuring the mood temperature or used to calibrate the overall mood temperature scale.

[0082] A Visitor Mood Pattern Learning Unit 240, which is used to collect and analyze the visitor's past mood data, learn the visitor's mood pattern, and predict the possible mood changes of the visitor.

[0083] Figure 4 This is the structural block diagram of the audio-visual material screening algorithm module provided by the embodiment of the present invention. As Figure 4 shown, the audio-visual material screening algorithm module specifically includes:

[0084] A Visitor Profile Construction Unit 310, which is used to collect the visitor's previous review data, including the types of audio-visual materials reviewed, review time, review frequency, analyze the visitor's preferences and habits, and construct a customized profile for the visitor;

[0085] A Screening Strategy Determination Unit 320, which is used to determine the content categories and types to be screened according to the visitor profile, environmental feature score, and mood temperature score. The specific formula is as follows:

[0086] [R(U, V) = p1×H_{user} + p2×Y_{user} + p3×P(U, V) + p4×C(U, V) + \delta], where R(U, V) is the score determined by the screening strategy, U is the visitor, V is the candidate audio-visual material, H_{user} is the feature score of the visitor's current environment, from the environment and time recognition module, Y_{user} is the mood temperature of the visitor, from the visitor mood perception module, p1, p2, p3, p4 are weight parameters, \delta is a constant bias term, P(U, V) is the score calculated for the preference similarity between the previously reviewed data of visitor U and candidate audio-visual material V, and C(U, V) is the content analysis score, according to the matching degree between the content of the audio-visual material and the visitor's preference;

[0087] The screening content matching unit 330, for which the goal of the algorithm is to maximize the R(U, V) score, that is, the expected satisfaction of visitor U with audio-visual material V, and screens a series of audio-visual materials with the highest scores for the visitor.

[0088] Figure 5 The structural block diagram of the visitor feedback learning module provided by the embodiment of the present invention is as Figure 5 shown, and the visitor feedback learning module specifically includes:

[0089] The feedback learning optimization unit 410, which is used to monitor the visitor's review completeness, review time, and the reaction of the visitor's score to the screened content, analyze the visitor's feedback, and dynamically adjust and optimize the visitor profile and screening strategy;

[0090] The visitor interaction and feedback unit 420, which is used to provide a visitor interface, display the screened audio-visual materials, collect direct feedback from the visitor on liking, favoriting, and commenting on the screened content, and further optimize the visitor profile and screened content through the visitor's interaction behavior.

[0091] Figure 6 The structural block diagram of the visitor interface and interaction module provided by the embodiment of the present invention is as Figure 6 shown, and it further includes the visitor interface and interaction module, which specifically includes:

[0092] The audio-visual material screening display unit 510, which is used to display the list of screened audio-visual materials;

[0093] The visitor customized setting unit 520, which is used to provide personal information management, view and edit personal profiles, set the preference adjustment function, and specify and adjust the weights of factors in the screening algorithm.

[0094] Figure 7 The structural block diagram of the visitor privacy protection module provided by the embodiment of the present invention is as Figure 7 shown, and it further includes the visitor privacy protection module, which specifically includes:

[0095] A visitor consent and data collection unit 610, configured to display the purpose and scope of data collection through a visitor interface and obtain collection authorization before using any data collection function;

[0096] A data security and encryption unit 620, configured to encrypt all collected data.

[0097] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown in sequence according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limitation, and these steps can be executed in other orders. Moreover, at least a part of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0098] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0099] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as falling within the scope described in this specification.

[0100] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

[0101] The foregoing is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A personalized video recommendation system based on user usage habits, characterized in that, The system includes: An environment and time recognition module, which is used to obtain the location information of the visitor through the GPS of the portable multimedia player, read the system time, and collect environmental volume and light intensity data through the microphone and sensor of the portable multimedia player to assist in determining the environment where the visitor is currently located; A visitor mood perception module, which is used to capture the dynamic facial features of the visitor through the front camera, capture the body language of the visitor, determine the state of the visitor, and combine the data of the dynamic facial features and body language to create the mood temperature state of the visitor; An audio-visual material screening algorithm module, which is used to combine the previous review data of the visitor and the current mood temperature state to construct a customized visitor profile. According to the visitor profile and the current mood temperature state, and integrating location, time, and environmental factors into the screening logic, it matches the content that meets the screening logic from the audio-visual material library; A visitor feedback learning module, which is used to monitor the visitor's review of the screened audio-visual materials, record the completeness of the visitor's review of the screened audio-visual materials, optimize the visitor profile and screening algorithm, and provide a feedback mechanism for the visitor.

2. The system according to claim 1, wherein The environment and time recognition module specifically includes: A location positioning unit, which is used to obtain the current geographical coordinates of the visitor using the GPS function of the portable multimedia player, and combine the GPS data with the geographical information system to judge the macro environment where the visitor is located, recorded as spatial location data; A time acquisition unit, which is used to directly obtain the current time from the system clock of the device and analyze which period of the day the visitor is in, recorded as time period data; An environment perception unit, which is used to measure the noise level of the surrounding environment using the microphone of the device, recorded as noise level data, and use the light sensor and camera to evaluate the light intensity of the current environment, recorded as light intensity data; A visitor environment determination unit, which integrates the spatial location data, time period data, noise level data, and light intensity data to determine the characteristics of the specific environment where the visitor is located. The specific formula is: H = w1×S + w2×T + w3×N + w4×L H is the environmental characteristic score, S is the spatial location data, T is the time period data, N is the noise level data, L is the light intensity data, and w1, w2, w2, w3 are weight factors.

3. The system according to claim 1, wherein The visitor mood perception module specifically includes: A facial feature recognition unit, which is used to use the front camera of the device to capture the facial feature image of the visitor in real time, locate the facial features of the visitor, analyze the facial feature data, and identify the basic mood represented by the visitor's features, recorded as facial feature mood data; A body language analysis unit, which is used to capture the body movements and postures of the visitor, identify the key points of the body, analyze the positions and movements of the body key points, and identify the body language of the visitor, recorded as body language mood data; A mood temperature calculation unit, which is used to calculate the mood intensity of the visitor based on the facial feature mood data and body language mood data, recorded as the mood temperature score. The specific formula is: Y = ɑ×E + β×B + γ Y is the modal temperature score, α and β are weight factors, E is the modal data of facial feature characteristics, B is the modal data of body language, and γ is the baseline offset; The visitor modal pattern learning unit is used to collect and analyze the past modal data of visitors, learn the modal patterns of visitors, and predict the possible modal changes of visitors.

4. The system according to claim 3, wherein The audio-visual material screening algorithm module specifically includes: The visitor profile construction unit is used to collect the previous review data of visitors, including the types of audio-visual materials reviewed, review time, review frequency, analyze the preferences and habits of visitors, and construct a customized profile for visitors; The screening strategy determination unit is used to determine the content categories and types to be screened according to the visitor profile, environmental feature score, and modal temperature score. The specific formula is: [R(U, V) = p1×H_user + p2×Y_user + p3×P(U, V) + p4×C(U, V) + δ] Among them, R(U, V) is the screening strategy determination score, U is the visitor, V is the candidate audio-visual material, H_user is the feature score of the visitor's current environment, from the environment and time recognition module, Y_user is the modal temperature of the visitor, from the visitor modal perception module, p1, p2, p3, p4 are weight parameters, δ is a constant bias term, P(U, V) is the preference similarity calculation score between the previous review data of visitor U and candidate audio-visual material V, and C(U, V) is the content analysis score, according to the matching degree between the content of the audio-visual material and the visitor's preferences; The screening content matching unit is used to maximize the R(U, V) score, that is, the expected satisfaction of visitor U with audio-visual material V, and screen a series of audio-visual materials with the highest scores for the visitor.

5. The system according to claim 4, wherein The visitor feedback learning module specifically includes: The feedback learning optimization unit is used to monitor the review completeness, review time, and the reaction of the visitor's score to the screened content of the visitor, analyze the visitor's feedback, and dynamically adjust and optimize the visitor profile and screening strategy; The visitor interaction and feedback unit is used to provide a visitor interface, display the screened audio-visual materials, collect the direct feedback of the visitor on the screened content, and further optimize the visitor profile and screened content through the interaction behavior of the visitor.

6. The system according to claim 1, wherein It also includes a visitor interface and interaction module, which specifically includes: The audio-visual material screening display unit is used to display the list of screened audio-visual materials; The visitor customized setting unit is used to provide personal information management, view and edit personal profiles, set preference adjustment functions, and specify and adjust the weights of factors in the screening algorithm.

7. The system according to claim 1, wherein It also includes a visitor privacy protection module, which specifically includes: The visitor consent and data collection unit is used to display the purpose and scope of data collection through the visitor interface before using any data collection function, and obtain collection authorization; The data security and encryption unit is used to encrypt all collected data.

Images