Family time-sharing portrait personality recommendation method and device based on real-time detection, and terminal

Abstract

The invention discloses a family time-sharing portrait personality recommendation method and device based on real-time detection and a terminal. The method comprises the following steps: acquiring watching behavior data related to a user watching behavior in real time; performing content real-time detection on the watching behavior data, and judging preference tendencies of a current watching user to different types of contents; performing feature extraction on operation behavior data in the watching behavior data, and determining an operation behavior mode of the current watching user; monitoring watching environment data of the intelligent terminal, and assisting in judging identity information and watching requirements of a current watching user; generating a user portrait of the current watching user in each time period according to the judged preference tendency, the operation behavior mode, the identity information and the watching demand of the current watching user in combination with a preset time period division rule; according to the generated user portrait of the current watching user in each time period, personalized recommendation content is automatically provided for the current watching user, recommendation and display are carried out, and convenience is provided for use of the user.

Description

Technical Field

[0001] The present invention relates to the field of smart TV recommendation technology, and in particular to a method, device, smart terminal and storage medium for implementing personalized recommendation of family time-sharing portraits based on a real-time detection engine. Background Art

[0002] With the popularity of smart TVs, TV, as a large public screen in the home, faces an important problem: due to the diversity of family users, different members may watch TV at different times, but it is difficult for the TV system to accurately judge the behavioral portrait of the user watching at that time.

[0003] Traditional recommendation methods in the prior art are often based on the viewing history and static data of the entire family, and cannot make timely adjustments based on the user's real-time viewing behavior, cannot meet the needs of family users for personalized content at different times, and cannot make accurate recommendations for specific viewing users. This has resulted in many personalized recommendations failing to effectively reach actual viewing users, affecting user experience and reducing the value and effectiveness of TV content recommendations.

[0004] Therefore, the existing technology still needs to be improved and developed. Summary of the invention

[0005] The technical problem to be solved by the present invention is that, in view of the problems and defects of the above-mentioned prior art, a method, device, intelligent terminal and storage medium for personalized recommendation of family time-sharing portrait based on real-time detection are provided. The present invention solves the problem that the television cannot accurately judge the current viewing user behavior portrait, resulting in inaccurate personalized recommendation; and provides a method for accumulating time-sharing family portraits of family television users by using a real-time detection engine, and then realizing user time-sharing personalized recommendation, thereby improving user experience and the effectiveness of television content recommendation, and providing convenience for users.

[0006] The technical solution adopted by the present invention to solve the problem is as follows: A method for recommending personality of family time-sharing portrait based on real-time detection, comprising: The intelligent terminal collects and obtains viewing behavior data related to the user's viewing behavior in real time; Performing real-time content detection on the viewing behavior data collected in real time, extracting video metadata and picture features from the video content data in the viewing behavior data, and determining the preference of the current viewing user for different types of content; Performing real-time behavior detection on the viewing behavior data collected in real time, extracting features from the operation behavior data in the viewing behavior data, classifying and pattern-recognizing the extracted behavior features, and determining the operation behavior pattern of the current viewing user; At the same time, it conducts real-time environmental detection, monitors the viewing environment data of smart terminals, and assists in determining the identity information and viewing needs of the current viewing user; Generate a user profile for each time period of the current viewing user based on the determined preference, operation behavior pattern, identity information, and viewing needs of the current viewing user and the preset time period division rules; Based on the generated user portrait of the current viewing user in each time period, personalized recommended content is automatically provided to the current viewing user and displayed.

[0007] The method for recommending family time-sharing portraits based on real-time detection, wherein the step of generating a user portrait for each time period of the current viewing user based on the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user in combination with a preset time period division rule further includes: Establish a real-time feedback mechanism to continuously input new real-time detection data into the time-sharing portrait generation model, recalculate the current user's preference weight for content in the corresponding time period, and adjust the user portrait.

[0008] In the method for recommending family time-sharing portrait personality based on real-time detection, the step of the smart terminal acquiring viewing behavior data related to the user's viewing behavior in real time comprises: The viewing behavior data related to the user's viewing behavior includes: viewing behavior data related to the user's active viewing behavior, and viewing behavior data related to randomly sending different content types to the user to test whether the user has a click preference for this type of content; Utilize multiple sensors and software monitoring mechanisms built into smart terminals to collect and obtain data related to user viewing behavior in real time; the viewing behavior data includes user operation data on the remote control, real-time feedback data of the terminal screen, voice command data of the user's interaction with the terminal, and application opening and usage data; The collected viewing behavior data is initially processed by data format conversion and / or data compression.

[0009] The method for personalized recommendation of family time-sharing portrait based on real-time detection, wherein the steps of performing real-time content detection on the viewing behavior data collected in real time, extracting video metadata and picture features from the video content data in the viewing behavior data, and judging the preference tendency of the current viewing user for different types of content include: Analyzing and preprocessing the viewing behavior data collected in real time; For video content data, extract the metadata and picture features of the video, and combine it with the user's operation behavior data on the video content, and use the content-based analysis algorithm to calculate the current user's real-time interest value for different types of content; Among them, the metadata includes data such as title, type and / or tag, the picture features include color distribution features and / or scene type features, and the operation behavior data includes viewing time data, whether to pause behavior data, and whether to fast-forward behavior data.

[0010] The method for recommending family time-sharing portrait personality based on real-time detection, wherein the steps of performing real-time detection on the viewing behavior data collected in real time, extracting features from the operation behavior data in the viewing behavior data, classifying and pattern-recognizing the extracted behavior features, and determining the operation behavior pattern of the current viewing user include: Using a behavior analysis algorithm, performing sequence analysis and pattern recognition on the collected operation data of the viewing behavior data, analyzing the time sequence of user operations, and determining the rhythm and regularity of user operations; The machine learning model is used to classify and predict the user's operating behavior patterns, and determine the operating behavior pattern of the current viewing user.

[0011] The method for recommending family time-sharing portrait personality based on real-time detection, wherein the step of generating a user portrait for each time period of the current viewing user according to the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user, combined with the preset time period division rule, comprises: Obtain the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user, as well as the preset time period division rules; A combination of rule-based and machine learning methods is used. In terms of rule-based, initial portrait templates for different time periods are set; A machine learning algorithm is used to train and optimize the initial portrait template using historical data to obtain the probabilistic relationship between different detection results and user portrait features in each time period, so that the initial portrait template can be dynamically adjusted according to real-time detection data to obtain the user portrait of the current viewing user in each time period.

[0012] The method for personalized recommendation of family time-sharing portraits based on real-time detection, wherein the steps of automatically providing personalized recommended content to the current viewing user based on the generated user portraits of each time period of the current viewing user and recommending and displaying the content include: Provide personalized recommended content for the current viewing user based on the user portrait of each time period of the current viewing user; the recommended content includes video programs, applications, and advertising products, and the recommended content is dynamically adjusted according to the changes in the user's time-sharing portrait in different time periods; The recommended content is displayed in real time.

[0013] A family time-sharing portrait personality recommendation device based on real-time detection, wherein the device comprises: A real-time data collection module, used to control the intelligent terminal to collect and obtain viewing behavior data related to the user's viewing behavior in real time; The content real-time detection submodule is used to perform real-time content detection on the viewing behavior data collected in real time, extract video metadata and picture features from the video content data in the viewing behavior data, and determine the preference of the current viewing user for different types of content; A behavior real-time detection submodule is used to perform behavior real-time detection on the viewing behavior data collected in real time, extract features of the operation behavior data in the viewing behavior data, classify and pattern recognize the extracted behavior features, and determine the operation behavior pattern of the current viewing user; The real-time environment detection submodule is used to simultaneously perform real-time environment detection, monitor the viewing environment data of the smart terminal, and assist in determining the identity information and viewing needs of the current viewing user; The family time-sharing portrait module is used to generate a user portrait for each time period of the current viewing user based on the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user, combined with the preset time period division rules; The real-time portrait update submodule is used to establish a real-time feedback mechanism, continuously input new real-time detection data into the time-sharing portrait generation model, recalculate the current user's preference weight for content in the corresponding time period, and adjust the user portrait; A personalized recommendation engine module is used to automatically provide personalized recommended content to the current viewing user based on the generated user portraits of the current viewing user in each time period; The recommended content presentation module is used to display personalized recommended content.

[0014] An intelligent terminal includes a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, including the method for executing any one of the methods described above.

[0015] A computer-readable storage medium, wherein when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute any one of the methods described above.

[0016] Beneficial effects of the present invention: The present invention provides a method, device, intelligent terminal and storage medium for personalized recommendation of family time-sharing portraits based on real-time detection. The present invention continuously collects and obtains user viewing behavior data; analyzes the collected user viewing behavior data respectively, generates time-sharing portraits according to the real-time detection results, and continuously updates the time-sharing portraits according to subsequent real-time detection data, and provides the real-time time-sharing portraits to the personalized recommendation engine module; the personalized recommendation engine module generates personalized recommendation content based on the real-time time-sharing portraits, and passes these contents to the recommended content presentation module. The recommended content presentation module displays the personalized recommended content to the user, collects the user's feedback on the recommended content (such as clicks, viewing time, etc.), and feeds the feedback data back to the real-time data collection module for further optimizing the operation of each module.

[0017] The present invention can make timely adjustments based on the user's real-time viewing behavior, can promptly meet the needs of home users for personalized content at different time periods, and can make accurate recommendations for specific viewing users; it realizes that many personalized recommendations can effectively reach actual viewing users, improves user experience, and also improves the value and effect of TV content recommendations, providing convenience for users. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flowchart of a method for recommending personality based on family time-sharing portraits based on real-time detection provided in Example 1 of the present invention.

[0020] Figure 2 It is a schematic diagram of the processing flow of the family time-sharing portrait personality recommendation method based on real-time detection provided in Example 2 of the present invention.

[0021] Figure 3 A principle block diagram of an embodiment of a family time-sharing portrait personality recommendation device based on real-time detection provided by the present invention.

[0022] Figure 4 It is a block diagram of the internal structure principle of the intelligent terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. 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.

[0024] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0025] Traditional recommendation methods in existing technologies are often based on the viewing history and static data of the entire family. They cannot make timely adjustments based on the user's real-time viewing behavior, cannot meet the needs of family users for personalized content at different times, and cannot make accurate recommendations for specific viewing users. This has led to many personalized recommendations failing to effectively reach actual viewing users, affecting user experience and reducing the value and effectiveness of TV content recommendations.

[0026] The present application provides a method for implementing personalized recommendation of household time-sharing portraits based on a real-time detection engine, which solves the problem that the television cannot accurately judge the current viewing user behavior portrait, resulting in inaccurate personalized recommendations, and improves user experience and the effectiveness of television content recommendations.

[0027] like Figure 1 As shown, a method for recommending personality of a family time-sharing portrait based on real-time detection in Embodiment 1 of the present invention comprises the following steps: Step S100: The smart terminal collects and obtains viewing behavior data related to the user's viewing behavior in real time; The smart terminal described in the embodiment of the present invention is described by taking a smart TV as an example. The smart TV can collect and obtain viewing behavior data related to the user's viewing behavior in real time.

[0028] In specific implementation, high-precision operation sensors (such as accelerometers and gyroscopes for remote control operation monitoring), ambient light sensors, microphones and other hardware devices can be installed on smart TV terminals.

[0029] At the operating system level of smart TVs, a data collection program is developed, which obtains hardware sensor data and software operation data in a low-latency and high-frequency manner. For example, through the event monitoring mechanism of the operating system, every key operation of the user on the remote control is captured, and the timestamp of the operation, key type and other information are recorded.

[0030] Each hardware sensor can convert physical signals into electrical signals, and the software monitoring mechanism obtains the operation information of the application layer through the underlying interface of the operating system. These data are collected at very small time intervals (such as every 0.5 seconds) and are preliminarily processed and packaged through the network module of the smart TV.

[0031] The specific process is to use the multiple sensors and software monitoring mechanisms built into smart TVs to capture viewing behavior data related to user viewing behavior in real time. These data include but are not limited to the user's operation data on the remote control (such as key frequency, operation type, and operation time interval), real-time feedback data of the TV screen (such as the channel watched, the type of content played, and the screen stay time), and other information about the user's interaction with the TV (such as voice commands, the opening and use of applications).

[0032] Then, the present invention will perform preliminary processing on the collected viewing behavior data related to the user's viewing behavior, such as data format conversion, data compression, etc., to reduce the network transmission burden. At the same time, a reliable network transmission protocol (such as a custom protocol based on UDP) is used to send the data to the real-time detection engine module in the background in real time. In order to ensure the stability of data transmission, a data cache and retransmission mechanism is set to cache data when the network is unstable and retransmit it in time after the network is restored.

[0033] In the specific implementation of the present invention, the viewing behavior data related to the user's viewing behavior includes: viewing behavior data related to the user's active viewing behavior, and viewing behavior data related to randomly sending different content types to the user to test whether the user has a click preference for this type of content; Among them, viewing behavior data related to user's active viewing behavior: This part of the data refers to the situation where users spontaneously choose to watch content. For example, users actively click on a video, article or other media content. These behaviors directly reflect the user's interests and preferences.

[0034] Collecting this type of data can provide a deep understanding of users' viewing habits and preferences, help the platform optimize content recommendations, and provide a personalized user experience. For example, if a user often watches technology videos, the system will record this behavior and push more technology-related content to increase user stickiness.

[0035] Different content types are randomly sent to users to test whether they have a click preference for this type of content. This part of the data is obtained by randomly pushing different types of content to users (for example, funny videos, practical tutorials, news reports, etc.) and observing their click behavior to determine their preferences for these content types.

[0036] For example, when a user does not usually watch cooking videos, but randomly pushes a popular food video, if the user clicks to watch and interacts (such as commenting or liking), the platform will take the user's cooking interests into consideration and may recommend more food content in the future. The present invention can obtain more comprehensive data on user preferences through this A / B testing or experimental method, thereby identifying and mining potential content interests and optimizing content classification and recommendation algorithms.

[0037] In this way, the embodiment of the present invention can help to more accurately understand user preferences by combining these two types of data, thereby achieving personalized recommendations. And by analyzing user behavior, it can provide content that better suits user interests and improve the user's viewing experience.

[0038] Step S200: performing real-time content detection on the viewing behavior data collected in real time, extracting video metadata and picture features from the video content data in the viewing behavior data, and determining the preference of the current viewing user for different types of content; In this step, the viewing behavior data collected in real time is analyzed in real time, and the video metadata and picture features are extracted from the video content data in the viewing behavior data to determine the current viewing user's preference for different types of content. For example, when the user switches channels or browses content, the user's interest in news, sports, movies, children's programs, etc. can be quickly determined.

[0039] Specifically, a content-based real-time analysis algorithm can be used to quantify the user's interest in different content types by identifying the category of the currently playing content (such as through video metadata, picture feature analysis, etc.) and the user's operation behavior on the content (such as viewing time, whether actively searching, etc.). The calculation of the interest value comprehensively considers factors such as the exposure time of the content and the user's interaction frequency.

[0040] In the embodiment of the present invention, after the background receives the viewing behavior data collected in real time, the data is first parsed and preprocessed. For video content data, the metadata (such as title, type, tag) and picture features (such as color distribution, scene type) of the video are extracted. And combined with the user's operation behavior data on the content (such as viewing time, whether to pause, whether to fast forward, etc.), the content-based analysis algorithm is used to calculate the current user's real-time interest value for different types of content (i.e., quantitative algorithm). For example, a content-user interest matrix is ​​used to update the interest value in the matrix according to the length of time the user watches a certain type of content and the frequency of operation.

[0041] It can be seen that the content-based real-time analysis method in the embodiment of the present invention identifies content categories by extracting video metadata and picture features, and at the same time combines the user's operational behavior data on the content, and uses a specific algorithm (such as the content-user interest matrix) to quantify the user's real-time interest value for different types of content; the specific technical method for content category identification, interest value quantification algorithm and its parameter setting, and the fusion analysis method of operational behavior data and content data adopted have changed the traditional method of judging content preferences based on historical data, realized real-time detection of users' current content preferences, and realized more accurate real-time detection of preferences.

[0042] Specifically, the step S200 includes: S201, parsing and preprocessing the viewing behavior data collected in real time; In this step, when users watch videos and other content, various viewing behavior data of users will be collected in real time, such as the time of viewing, viewing duration, and operations performed (such as pause, fast forward, etc.). The parsing and preprocessing step is to clean and convert these raw data. For example, some erroneous or incomplete data records are removed, and the data format is unified for more accurate subsequent analysis.

[0043] S202: extracting metadata and picture features of the video content data, and combining the user's operation behavior data on the video content, using a content-based analysis algorithm to calculate the current user's real-time interest value for different types of content; Among them, the metadata includes data such as title, type and / or tag, the picture features include color distribution features and / or scene type features, and the operation behavior data includes viewing time data, whether to pause behavior data, and whether to fast-forward behavior data.

[0044] Among them, metadata and picture features are extracted: metadata is some basic information about the video, such as the title can directly reflect the theme of the video, and the type (such as movie, documentary, animation, etc.) and tag data (such as action, science fiction, comedy, etc.) further refine the attributes of the video. In terms of picture features, color distribution features can reflect the visual style of the video. For example, a video with an overall warm tone may give people a warm feeling; scene type features can determine whether the video is indoors, outdoors, or a battle scene, a dialogue scene, etc.

[0045] The embodiment of the present invention calculates the interest value in combination with the operation behavior data. Specifically, the user's operation behavior data can directly reflect his reaction to the video content. Long viewing time data indicates that the user is more interested in this part of the content; pause behavior may mean that the user is paying attention to a detail or thinking; fast forward behavior may indicate that the user is not interested in the current content. The content-based analysis algorithm will integrate this information, quantify the user's interest in different types of content (such as action type, comedy type, etc.) at the current moment, and obtain a real-time interest value.

[0046] For example, when a user is watching content on a video platform, in step S201, the present invention collects the user's viewing behavior data and finds that the viewing time format of one record is wrong. After parsing and preprocessing, it is corrected to the correct format.

[0047] In step S202, the metadata of a video shows that the title is "Thrilling Wild Adventure", the type is "Adventure", and the tags are "Wild", "Adventure", "Challenge", etc. In terms of picture features, the color distribution is mainly green (representing the wild environment) and brown (representing the terrain), and the scene types are mainly wild walking and exploration scenes. When the user watched this video, the viewing time reached 15 minutes (relatively long), during which there was no pause or fast-forward behavior. At the same time, the user also watched another comedy-type video, but the viewing time was only 3 minutes and there were multiple fast-forward behaviors. Based on these data, it is calculated through a content-based analysis algorithm that the user has a higher real-time interest value for adventure-type content, and a lower real-time interest value for comedy-type content. The present invention can subsequently recommend more similar adventure-type video content to users based on this interest value.

[0048] It can be seen that by accurately calculating the real-time interest value of users, the present invention can more accurately recommend content that users may be interested in, reduce the time and energy of users in finding content of interest, and improve users' satisfaction with the platform and frequency of use. Based on the in-depth understanding of user interests, the present invention can optimize the recommendation algorithm, improve the accuracy and relevance of recommendations, and enhance the overall recommendation effect, thereby increasing the user's stay time and interaction on the platform.

[0049] Step S300: performing real-time behavior detection on the viewing behavior data collected in real time, extracting features from the operation behavior data in the viewing behavior data, classifying and pattern-recognizing the extracted behavior features, and determining the operation behavior pattern of the current viewing user; In the embodiment of this step, the viewing behavior data collected in real time is detected in real time to analyze the operation behavior pattern of the current viewing user, including the fluency of the operation, the complexity of the operation, the continuity of the operation, etc. For example, it is determined whether the user browses the channel quickly or watches for a long time, whether the user is proficient in the operation, whether there is repeated operation, etc.

[0050] Specifically, the behavior analysis algorithm can be used to perform sequence analysis and pattern recognition on the collected operation data of the viewing behavior data. For example, by analyzing the time series of user operations, the rhythm and regularity of user operations can be determined, and machine learning models (such as hidden Markov models) can be used to classify and predict user operation behavior patterns.

[0051] For example, when there is a video platform, users watch various video contents on the platform. The viewing behavior data collected in real time by the present invention includes the user's video viewing time, viewing duration, pause, fast forward, fast rewind, skip advertisement and other operation behaviors.

[0052] When performing real-time behavior detection, the present invention will continuously monitor every operation of the user. For example, when a user is watching a cooking instruction video, he suddenly clicks the pause button several times in a row. The present invention immediately captures this abnormal pause operation behavior, which is real-time behavior detection.

[0053] Regarding the extraction of operational behavior data features, the present invention extracts relevant features for the user's multiple quick clicks on pause just now. For example, within just 10 seconds, the number of pause operations reached 5 times, and each pause time was extremely short, with an average of only 0.5 seconds per pause. These data, such as the pause frequency, the duration of each pause, etc., are the extracted behavioral features.

[0054] Then, the behavioral feature classification and pattern recognition are performed: the present invention compares the extracted features with the preset behavioral pattern library. It is found that this high-frequency and short-term pause operation meets the behavioral pattern characteristics of "paying attention to video details". In the behavioral pattern library, multiple behavioral patterns such as "casual browsing", "focused learning", "looking for specific clips" and their corresponding features have been defined. Through this comparison, the system determines the current user's operation behavior pattern to determine whether the user is browsing quickly or watching attentively.

[0055] That is, in the embodiment of the present invention, the user operation behavior data is feature extracted, including operation time interval, operation speed, operation path (such as the order of remote control button operation), etc. Then, these behavior features are classified and pattern recognized using machine learning models (such as decision trees and neural networks). For example, a decision tree model is trained to determine whether the user is browsing quickly or watching attentively based on the operation time interval and operation speed.

[0056] In this way, the present invention can better understand user needs by determining the user's operation behavior pattern. For example, if it is recognized that the user is "paying attention to video details", relevant video tips can be automatically popped up for later real-time personalized recommendations, or text explanations of the details can be provided to help users better understand the video content and improve the user's viewing experience.

[0057] And understand the user's operation behavior pattern, the present invention can make video recommendations more accurately. If a user often shows the behavior pattern of "focusing on learning", the present invention can recommend more high-quality knowledge videos in the subsequent time-sharing recommendation, improve the accuracy and relevance of the recommendation, and increase user stickiness. The feedback data is fed back to the real-time data acquisition module to further optimize the operation of each module.

[0058] It can be seen that the real-time behavior detection in the embodiment of the present invention extracts features from user operation behavior data, including operation time interval, speed, path, etc., and then uses machine learning models (such as hidden Markov models, decision trees, etc.) for classification and pattern recognition, thereby analyzing user operation behavior patterns in real time; it solves the problem that traditional technologies cannot understand user operation behavior habits in real time.

[0059] Step S400: Real-time environmental detection is performed simultaneously to monitor the viewing environment data of the smart terminal, and to assist in determining the identity information and viewing needs of the current viewing user; In the embodiment of the present invention, real-time environmental detection is also performed to monitor data related to the smart TV viewing environment, such as viewing time (accurate to minutes), ambient light intensity, ambient noise level, etc. These environmental factors can assist in determining the possible identity and viewing needs of the current viewing user.

[0060] Specifically, the present invention can measure the light intensity through the ambient light sensor set in the smart TV, collect the ambient noise data through the microphone, and provide the accurate viewing time through the clock module of the smart TV. Through the comprehensive analysis of these data, for example, in the time period of dim light and quiet environment, it is speculated that it may be adults watching leisurely in the evening; in the case of bright light and certain noise, it may be that many people in the family are watching during the day.

[0061] Specifically, the ambient light sensor data can be calibrated and quantified to obtain the ambient light intensity value. The sound data collected by the microphone can be filtered and analyzed to obtain the ambient noise level. Combined with the precise time provided by the system clock of the smart TV, an environmental data vector is formed.

[0062] By analyzing historical environment data and corresponding user portrait data, we can establish an environment-user portrait mapping relationship. For example, through a large amount of data statistics, we found that during the period of time when the light is dim and the environment is quiet, most users are adults who watch leisurely in the evening. Based on this, we can infer the user portrait of the current environment data to assist in determining the identity information and viewing needs of the current viewing user.

[0063] It can be seen that the real-time environment detection in the embodiment of the present invention quantifies and analyzes the data collected by the ambient light sensor and microphone and the time provided by the system clock, and combines historical data statistics to establish an environment-user portrait mapping relationship, thereby assisting in determining the identity and needs of the current viewing user, thereby making up for the deficiency of traditional recommendation systems that ignore the impact of the viewing environment on user behavior.

[0064] Step S500: Generate a user profile for each time period of the current viewing user based on the determined preference, operation behavior pattern, identity information and viewing needs of the current viewing user and in combination with a preset time period division rule; In the embodiment of this step, the preference tendency of the current viewing user determined in the previous step, such as whether the user likes to watch comedy, action or documentary videos; and operation behavior patterns, such as whether the user often pauses or fast-forwards when watching videos, or watches them from the beginning to the end; identity information, such as age, gender, occupation, etc.; and viewing needs, such as for entertainment and relaxation, or for learning knowledge, etc. This information is obtained by analyzing the user's previous viewing history, operation records, and personal information filled in.

[0065] In combination with the time period division rule, that is, the time period division rule preset in the embodiment of the present invention can be divided according to different time periods of the day (morning, noon, evening, etc.), different days of the week (weekdays, weekends, etc.), or even different seasons, etc. The behavior and needs of users may be different in each time period.

[0066] In the embodiment of the present invention, when generating a user portrait for each time period of the current viewing user, specifically, the above collected user information and time period division rules are integrated to generate a specific user portrait for the user in each time period. This portrait includes various characteristics and preferences of the user in the time period. For example, in the evening time period, the user portrait may show that the user likes to watch light comedy videos in the evening to relax, and often pauses to take notes and other operations.

[0067] For example, suppose there is a user on a video platform. By analyzing his historical viewing records and operation behaviors, it is found that: Preferences: I like science fiction and suspense movies and TV series, and occasionally watch some science and technology documentaries.

[0068] The operating behavior pattern is: they often pause to think while watching videos, watch the exciting clips repeatedly, and leave messages under the videos to discuss.

[0069] Identity information: 28 years old, male, programmer, needs to work during the day on weekdays.

[0070] The viewing needs are: people hope to relax and entertain themselves by watching videos on weekday evenings and weekends, and also hope to learn some new scientific and technological knowledge.

[0071] Generate user portraits based on preset time period division rules (weekday evenings, weekend daytimes, weekend nights, etc.): Weekday evenings: The user portrait shows that the user may want to relax by watching science fiction or suspense TV series after get off work. As a programmer, he is very tired at work, so he may be more focused when watching and pause to think about the plot. The platform can recommend popular science fiction and suspense dramas to him recently, and provide relevant content for plot analysis.

[0072] Weekend daytime: User portraits show that users may have more time. In addition to entertainment and relaxation, they may also watch some science and technology documentaries to learn new knowledge. The platform can recommend some high-quality science and technology documentaries and provide relevant learning materials and discussion communities.

[0073] Weekend nights: Users may prefer to watch light-hearted science fiction movies to relax, and may watch and discuss them with friends. The platform can recommend science fiction movies suitable for multiple people to watch, and provide online viewing and discussion functions.

[0074] In this way, according to the user portraits in different time periods, the embodiment of the present invention can more accurately recommend the content that users may be interested in in the time period. For example, on weekend afternoons, popular movies or variety shows can be recommended to users who like to relax and have fun, thereby improving the success rate of recommendations and user satisfaction.

[0075] After understanding the user's behavior patterns and needs in different time periods, the present invention can provide more personalized services. For example, during the time period when users often study at night, the present invention can provide them with recommendations for learning materials and tips on how to use learning tools, thereby enhancing the user's sense of dependence on the platform.

[0076] And by analyzing user portraits in different time periods, the present invention can understand the behavioral characteristics and demand changes of users in different time periods, so as to reasonably arrange server resources, optimize content update time, etc., and improve the overall operational efficiency of the platform.

[0077] That is, the embodiment of the present invention generates a time-sharing portrait of the current viewing user based on the real-time detection and analysis results and in combination with the time factor. The portrait content includes the user's preference for various types of content in different time periods, operation behavior characteristics, possible identity characteristics, etc.

[0078] A combination of rule-based and machine learning methods can be used. In terms of rule-based, set initial portrait templates for different time periods (such as weekday daytime, weekday evening, weekend daytime, and weekend evening); in terms of machine learning, use historical data to train and optimize these templates so that the portrait can be dynamically adjusted according to real-time detection data. For example, during the daytime on weekdays, if it is detected in real time that users frequently watch educational programs and operate relatively slowly, combine the portrait template of this time period to generate a time-sharing portrait of "weekday daytime may be for elderly people watching educational programs."

[0079] As can be seen from the above, the embodiment of the present invention generates a user portrait for each time period of the current viewing user based on the determined preference tendencies, operation behavior patterns, identity information and viewing needs of the current viewing user, combined with preset time period division rules, and by comprehensively considering the collaborative work of content, behavior, and environment, and integrating and interactively analyzing data results, a comprehensive and real-time detection result of the current viewing user is obtained, avoiding the limitations of single-dimensional detection and making the judgment of the user more accurate and comprehensive.

[0080] In specific implementation, the step S500 specifically includes: S501, obtaining the determined preference tendency, operation behavior pattern, identity information and viewing demand of the current viewing user, as well as the preset time period division rule; In this step, the user's past viewing records and operation behaviors (such as pause, fast forward, etc.) are analyzed to determine the user's preference (such as comedy or tragedy, action or art film, etc.) and operation behavior pattern (such as continuous viewing or frequent pause). At the same time, the user's identity information (such as age, gender, occupation, etc.) and the viewing needs inferred from user feedback or behavior (for entertainment, learning, socializing, etc.) are collected. In addition, the preset time division rules are obtained, such as dividing a day into different time periods such as morning, morning, afternoon, and evening, or dividing it according to weekdays and weekends in a week.

[0081] S502, using a method combining rule-based and machine learning, in terms of rule-based, setting initial portrait templates for different time periods; Based on the various information obtained in the first step, a combination of rule-based and machine learning methods is used. In the rule-based part, the initial user portrait template is set according to the characteristics of different time periods and the general behavior patterns of users. For example, on weekday mornings, considering that most users may be short on time, the initial portrait template set may be inclined to quickly browse short and interesting content, which may be some short video news or funny clips; on weekend nights, the template may be set so that users have more time, are more inclined to watch longer movies or TV series, and may have more social interaction needs.

[0082] S503: Use a machine learning algorithm to train and optimize the initial portrait template using historical data to obtain a probabilistic relationship between different detection results and user portrait features in each time period, so that the initial portrait template can be dynamically adjusted according to real-time detection data to obtain a user portrait of the current viewing user in each time period.

[0083] In an embodiment of the present invention, a large amount of historical data can be used to train the initial portrait template set previously through a machine learning algorithm. For example, the relationship between the actual viewing behavior of different types of users and their portrait features on weekday mornings in the past period of time is analyzed. Through this training, the probability relationship between different detection results (such as the user's specific operation behavior, viewing content type, etc.) and user portrait features (such as preferences, needs, etc.) is obtained. In this way, when there is real-time detection data, the initial portrait template can be dynamically adjusted according to these probability relationships and real-time data, thereby generating an accurate user portrait for the current viewing user in each time period. For example, if it is detected in real time that a user often clicks to watch short videos of science and technology on weekday mornings, then according to the probability relationship obtained by training, the user's portrait on weekday mornings can be adjusted to increase the weight of his preference for science and technology content.

[0084] For example, when a video platform has a user, who is 30 years old, male, and a designer, he is busy at work during the day and has more leisure time at night and weekends.

[0085] Stage S501: By analyzing the user's historical viewing records, it is found that he prefers design tutorial videos and some science fiction movies. His operating behavior pattern is that he frequently pauses to take notes when watching design tutorials, and occasionally fast-forwards to skip some uninteresting clips when watching movies. At the same time, the platform's preset time division rules are obtained, dividing a day into weekday mornings (7:00-9:00), weekday daytime (9:00-18:00), weekday evenings (18:00-23:00), weekend daytime (9:00-18:00), and weekend evenings (18:00-23:00).

[0086] Stage S502: Set the initial portrait template based on rules. On weekday mornings, considering that users may be preparing for work or commuting, the template is set that users may quickly browse some short design inspiration videos or news information. On weekday evenings, the template is set that users may watch design tutorial videos for learning and improvement, or watch science fiction movies to relax. During the daytime on weekends, the template is set that users have more time to learn design knowledge in depth, and may watch a longer series of design tutorials, and may also watch some science fiction documentaries. On weekend nights, the template is set that users are more inclined to entertainment and relaxation, may watch newly released science fiction movies, and may communicate and discuss with other users on the platform.

[0087] Stage S503: Use historical data for machine learning training. For example, by analyzing the viewing behavior data of designers similar to the user on weekday mornings over the past period of time, it is found that a certain proportion of designers watch short videos related to design trends in the morning. Through the machine learning algorithm, the probability relationship between the detection result of watching a short video on design trends on weekday mornings and the user's preference for design content is obtained. When it is detected in real time that the user clicks to watch a short video on design trends on a certain weekday morning, according to this probability relationship, the user's portrait on weekday mornings is adjusted to increase the weight of his preference for design content. At the same time, his viewing needs may be adjusted to obtain design inspiration, thereby further improving the user portrait in this time period.

[0088] It can be seen that in the embodiment of this step, by combining rule-based settings and machine learning optimization, known behavioral patterns and experience are used to set the initial template, and potential relationships are mined from a large amount of historical data through machine learning, so that the generated user portrait can more accurately reflect the user's real characteristics and needs.

[0089] The invention also has strong dynamic adaptability: it can dynamically adjust user portraits based on real-time detection data to adapt to the behavioral changes of different users in different time periods. For example, a user's interests may change with time, mood or life status, and this method can capture these changes in time and update the portrait.

[0090] The present invention can generate accurate user portraits for each time period of the current viewing user. Accurate user portraits can help the platform recommend content and services to users more accurately and meet the different needs of users in different time periods. For example, short and efficient content can be recommended in the user's busy time period, and richer and deeper content can be recommended in the leisure time period, thereby improving user satisfaction with the platform and frequency of use.

[0091] Step S600: Automatically provide personalized recommended content for the current viewing user based on the generated user portrait for each time period, and recommend and display the content.

[0092] In the embodiment of the present invention, highly personalized recommended content is provided to the current viewing user according to the user portrait of each time period of the current viewing user. The recommended content includes video programs, applications, advertising products, etc., and these recommended contents are dynamically adjusted according to the changes in the user's time-sharing portrait in different time periods.

[0093] Specifically, a combination of multiple recommendation algorithms can be used. For content recommendation, the content-based recommendation algorithm selects matching video programs based on the content preference characteristics in the user's time-sharing profile; the collaborative filtering recommendation algorithm looks for other users with similar time-sharing profiles to the current user and makes recommendations based on their viewing history. For application and advertising product recommendations, the user's behavior patterns and possible identity characteristics are comprehensively considered. For example, for time-sharing profiles with children's viewing characteristics, educational applications suitable for children and related children's product advertisements are recommended.

[0094] That is, in the embodiment of the present invention, the content-based recommendation algorithm can analyze the content preference characteristics in the time-sharing portrait and filter out matching video programs from the video content library. For example, if the time-sharing portrait shows that the user prefers action movies at night, action movies can be filtered out for recommendation based on the tags and classification information of the video content.

[0095] A collaborative filtering recommendation algorithm can be used to find other users with similar time-sharing profiles to the current user, and make recommendations based on their viewing history. For example, by calculating the similarity matrix of user time-sharing profiles, a user group with high similarity can be found, and popular action movies that they have watched but the current user has not watched can be recommended to the user.

[0096] For application and advertising product recommendations, the behavioral patterns and identity characteristics in the time-sharing portrait can be comprehensively considered. For example, for a time-sharing portrait with children's viewing characteristics, educational applications suitable for children can be selected from the application library for recommendation, and related products such as children's toys and school supplies can be selected from the advertising product library for advertising recommendations.

[0097] Then, in the embodiment of the present invention, the layout and display method of the television interface are designed according to the recommended content generated by the personalized recommendation engine, and the recommendation display is performed.

[0098] Specifically, for video program recommendations, popular recommended videos are displayed in the form of large icons and scrolling playback on the TV main interface. For example, a recommended action movie is displayed in the form of a high-definition poster in the center of the interface, and other recommended videos are displayed by scrolling.

[0099] For application recommendations, they are displayed in the form of a floating window or sidebar at the edge of the interface. For example, a semi-transparent floating window is displayed on the right side of the TV interface to recommend educational applications suitable for children, which is convenient for users to click and view.

[0100] For product recommendations, creative advertising that matches the user profile is displayed during the advertising period. For example, for users with elderly viewing characteristics, health care product advertisements are displayed during the advertising period. The advertising format uses concise and clear pictures and texts to match the visual habits of the elderly.

[0101] Based on user feedback, we continuously optimize the presentation of recommended content. For example, if users often ignore the application recommendations in the sidebar, we can adjust the position or presentation of the sidebar to improve the visibility and attractiveness of the recommended content.

[0102] In a further embodiment, the present invention also includes step S511, establishing a real-time feedback mechanism, continuously inputting new real-time detection data into the time-sharing portrait generation model, recalculating the current user's preference weight for content in the corresponding time period, and adjusting the user portrait.

[0103] That is, in the embodiment of the present invention, as the user's viewing behavior continues, the user's time-sharing portrait will be updated in real time to ensure that the portrait can accurately reflect the latest behavioral characteristics and preference changes of the current user.

[0104] In specific implementation, by establishing a feedback mechanism, the new data detected in real time is continuously input into the time-sharing portrait generation model, and the various feature parameters in the portrait are recalculated and adjusted. For example, if a user originally prefers to watch movies at night, but at a certain moment he starts to frequently switch to the sports channel and stay there for a long time, the portrait real-time update submodule will immediately adjust the user's portrait in the evening time period and increase his preference weight for sports content.

[0105] That is, in the embodiment of the present invention, by establishing a real-time feedback mechanism, new real-time detection data is continuously input into the time-sharing portrait generation model. For example, when a user suddenly watches sports channels for a long time in the evening when he originally preferred movies, this behavior data is input into the model, and the user's preference weight for sports content in the evening is recalculated to adjust the user profile.

[0106] The use of incremental learning algorithms enables the model to quickly adapt to new data changes without retraining the entire data set, ensuring the real-time and accuracy of time-sharing portraits.

[0107] The mf recognizes In another embodiment of the present invention, Figure 2 As shown, the overall architecture of the family time-sharing portrait personality recommendation method based on real-time detection of the present invention is realized. Figure 2The entire process from real-time data collection to the final presentation of recommended content is shown in the figure. Figure 2 Starting from the lower left corner of the figure, the real-time data collection module obtains viewing behavior data related to user viewing behavior, and then the data flows to the real-time detection engine module for multi-faceted detection and analysis. The analysis results then enter the family time-sharing portrait module to generate and update the time-sharing portrait, and then the personalized recommendation engine module generates recommended content based on the time-sharing portrait, and finally the recommended content presentation module presents it to the user. The modules are connected by arrows, which clearly indicate the flow direction of data and information, and reflect the coherence and collaborative working principle of the entire system.

[0108] Specifically, refer to Figure 2 As shown, a family time-sharing portrait personality recommendation method based on real-time detection in this specific application embodiment includes the following steps: S11. The real-time data collection module continuously transmits the collected user viewing behavior data to the real-time detection engine module.

[0109] S12. The content real-time detection submodule, the behavior real-time detection submodule and the environment real-time detection submodule in the real-time detection engine module analyze the data respectively and pass the analysis results to the family time-sharing portrait module.

[0110] In the embodiment of the present invention, a deep learning algorithm (such as a deep neural network) can be used to replace some existing detection algorithms. For example, in real-time content detection, a deep learning model can be used to perform more accurate content classification and user interest prediction on video images.

[0111] Specifically, a deep learning model can be trained by collecting a large amount of video content and user interaction data. The model can automatically learn the deep-level features of video content and the user's interest patterns in different content. For example, the convolutional neural network (CNN) is used to extract features from video frames, and combined with user viewing behavior data, the recurrent neural network (RNN) or long short-term memory network (LSTM) is used to predict the user's interest in the content. The deep learning model can mine more subtle features and relationships, thereby improving the accuracy of detection.

[0112] Of course, in the embodiment of the present invention, social network data can also be integrated for detection; that is, if the user authorizes, his social network data (such as the user's interests and hobbies on social media, topics of concern, etc.) can be integrated into the real-time detection engine to assist in determining the user's content preferences and behavior patterns.

[0113] Specifically, the user's account on the smart TV is associated with the social network account to obtain relevant data, which is converted into a feature vector and integrated with the data collected by the TV. For example, if the user often follows sports events on the social network, this information can be used as an important reference in the real-time detection of TV content to determine the user's potential interest in sports channels.

[0114] For users who are used to using social networks, this integration can further enrich the dimensions of user portraits and improve the accuracy of personalized recommendations. In particular, when it comes to content with social attributes (such as popular TV series, online topic programs, etc.), it can better match users' social interests.

[0115] S13. The time-sharing portrait generation submodule of the family time-sharing portrait module generates a time-sharing portrait according to the real-time detection results. The portrait real-time update submodule continuously updates the time-sharing portrait according to the subsequent real-time detection data, and provides the real-time time-sharing portrait to the personalized recommendation engine module.

[0116] In the embodiment of the present invention, based on the analysis results of the real-time detection engine and in combination with the preset time period division rules (such as weekday daytime: 6:00-18:00, weekday night: 18:00-24:00, etc.), the user portrait template is initialized for each time period. For example, the initial portrait template for weekday daytime may be "may be watched by the elderly or children, and the content preference is uncertain."

[0117] In specific implementation, machine learning algorithms (such as Bayesian networks) can be used to train historical data to obtain the probabilistic relationship between different detection results and user profile features in different time periods. For example, during the daytime on weekdays, if real-time content detection shows that users have a high interest in educational programs, and real-time behavior detection shows that operations are relatively slow, combined with the probabilistic relationship obtained through training, the user profile during the daytime on weekdays can be adjusted to "probably an elderly person watching educational programs."

[0118] Of course, when the present invention is implemented, in addition to dividing by time (such as daytime and night), time-sharing portraits can also be constructed according to the family life cycle (such as newlywed families, families with young children, empty-nest elderly families, etc.).

[0119] In this way, the user's registration information on the TV (such as age, family composition, etc.) and long-term viewing behavior data can be used to determine the life cycle stage of the family, and different portrait templates and weights can be formulated based on this. For example, a family with young children may prefer children's education and parent-child programs during the day on weekends, while a newlywed family may prefer romantic content in the evening.

[0120] This time-sharing portrait based on the family life cycle can more deeply explore the content needs of families at different stages and provide family users with more intimate and personalized recommendations that are more in line with their actual life conditions.

[0121] In addition, in the embodiment of the present invention, the time granularity of the time-sharing portrait is dynamically adjusted; the time granularity of the time-sharing portrait can be dynamically adjusted according to the user's viewing habits and behavior changes. For example, for users with more regular viewing time, a coarser time granularity (such as in hours) can be used; for users with irregular viewing time, a finer time granularity (such as in 15 minutes) can be used.

[0122] By analyzing the user's viewing time series data, the variance or regularity index of the viewing time is calculated, and the appropriate time granularity is determined based on the index value. For example, if the user watches TV regularly from 7 to 9 pm every day of the week, a coarser time granularity can be used during this time period; if the user's viewing time has no obvious regularity, a finer time granularity can be used throughout the day to build a time-sharing portrait. In this way, it can better adapt to the viewing habits of different users, improve the flexibility and accuracy of the time-sharing portrait, and thus improve the effect of personalized recommendations.

[0123] S14. The personalized recommendation engine module generates personalized recommendation content based on the real-time time-sharing portrait, and passes the content to the recommendation content presentation module.

[0124] In an embodiment of the present invention, in terms of a personalized recommendation engine, user emotion-aware recommendations can be introduced. Specifically, the user's current emotional state can be perceived through devices such as a camera or microphone of a smart TV, combined with emotional computing technology, and integrated into personalized recommendations.

[0125] The camera can be used to capture the user's facial expression, the microphone can be used to collect the user's voice tone, and the emotion recognition algorithm can be used to determine the user's emotions (such as happiness, sadness, anger, etc.), and then the recommended content can be adjusted according to the user's emotions and time-sharing portraits. For example, when the user is in a low mood, some light-hearted content can be recommended; when the user is in an excited mood, some exciting programs can be recommended.

[0126] In this way, it can better fit the user's current psychological state, provide more emotionally caring personalized recommendations, and enhance the user's viewing experience.

[0127] S15. The recommended content presentation module displays personalized recommended content to users, collects user feedback on the recommended content (such as clicks, viewing time, etc.), and feeds the feedback data to the real-time data collection module for further optimizing the operation of each module.

[0128] In the embodiment of the present invention, the recommended content presentation module presents the recommended content generated by the personalized recommendation engine module to the user in an intuitive and convenient manner. For example, the recommended video programs are highlighted on the TV interface, and application recommendations and advertising product recommendations are displayed in an appropriate manner to ensure that the user can easily discover and select the content of interest.

[0129] Specifically, the layout and display of recommended content can be optimized through the user interface design technology of smart TVs according to the user's operating habits and visual perception characteristics. For example, popular recommended videos can be displayed in a large icon, eye-catching colors, and scrolling playback, application recommendations can be displayed in the form of floating windows at the edge of the interface, and product recommendations can be displayed in the form of creative advertisements that match the user's profile during advertising periods.

[0130] In summary, the personalized recommendation method based on the real-time detection engine to realize the family time-sharing portrait can effectively solve the problem that the smart TV cannot accurately judge the current viewing user behavior portrait in real time, and significantly improve the user's viewing experience and the effect of TV content recommendation through real-time data collection, accurate detection, dynamic portrait and personalized recommendation. And the present invention also has the following advantages: 1) The time-sharing portrait generation and adjustment of the present invention combines time factors and uses a combination of rule-based and machine learning methods to generate and dynamically adjust the user's time-sharing portrait. First, set the time period division rules and the initial portrait template, then use machine learning algorithms (such as Bayesian networks) to train historical data to obtain the probabilistic relationship between detection results and portrait features, and finally dynamically adjust the portrait based on real-time detection data. This solves the problem that traditional user portraits cannot reflect the differences in user behavior in different time periods.

[0131] 2) The real-time updating of the portrait in the embodiment of the present invention creates a real-time updating method of the portrait based on the feedback mechanism and the incremental learning algorithm. As the user's viewing behavior continues, the new real-time detection data is continuously input into the time-sharing portrait generation model, and the incremental learning algorithm (such as the online learning algorithm) is used to quickly adapt to the data changes to achieve real-time updating of the portrait. This overcomes the defect of the traditional user portrait being updated in a timely manner.

[0132] 3) The method for dynamically adjusting the recommended content of the personalized recommendation engine of the present invention adopts a method for dynamically adjusting the recommended content according to the real-time changes of the user's time-sharing portrait. By obtaining the changes of the time-sharing portrait in real time and recalculating the recommended content, the timeliness and accuracy of the recommendation are ensured; the problem of the traditional recommendation system's fixed recommended content and inability to adapt to the real-time changing needs of users is avoided.

[0133] Exemplary Devices like Figure 3As shown, an embodiment of the present invention provides a family time-sharing portrait personality recommendation device based on real-time detection, and the device includes: The real-time data collection module 310 is used to control the intelligent terminal to collect and obtain viewing behavior data related to the user's viewing behavior in real time; The content real-time detection submodule 320 is used to perform real-time content detection on the viewing behavior data collected in real time, extract video metadata and picture features from the video content data in the viewing behavior data, and determine the preference of the current viewing user for different types of content; The behavior real-time detection submodule 330 is used to perform behavior real-time detection on the viewing behavior data collected in real time, extract features of the operation behavior data in the viewing behavior data, classify and pattern recognize the extracted behavior features, and determine the operation behavior pattern of the current viewing user; The real-time environment detection submodule 340 is used to simultaneously perform real-time environment detection, monitor the viewing environment data of the smart terminal, and assist in determining the identity information and viewing needs of the current viewing user; The family time-sharing portrait module 350 is used to generate a user portrait of each time period of the current viewing user according to the determined preference tendency, operation behavior pattern, identity information and viewing demand of the current viewing user and in combination with the preset time period division rules; The portrait real-time update submodule 360 ​​is used to establish a real-time feedback mechanism, continuously input new real-time detection data into the time-sharing portrait generation model, recalculate the current user's preference weight for content in the corresponding time period, and adjust the user portrait; The personalized recommendation engine module 370 is used to automatically provide personalized recommended content for the current viewing user based on the generated user portrait of the current viewing user in each time period; The recommended content presentation module 380 is used for displaying personalized recommended content, as described above in detail.

[0134] Based on the above embodiments, the present invention further provides an intelligent terminal, whose principle block diagram can be shown as follows: Figure 4 As shown. The intelligent terminal includes a processor, a memory, a network interface, a display screen, and a database connected through a system bus. Among them, the processor of the intelligent terminal is used to provide computing and control capabilities. The memory of the intelligent terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the intelligent terminal is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for recommending a personalized family time-sharing portrait based on real-time detection is implemented. The database of the intelligent terminal is used to store a personalized family time-sharing portrait recommendation program based on real-time detection.

[0135] Those skilled in the art will understand that Figure 4 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the scheme of the present invention, and does not constitute a limitation on the smart terminal to which the scheme of the present invention is applied. The specific smart terminal may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0136] In one embodiment, a smart terminal is provided, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, and the one or more programs include instructions for performing the following operations: The intelligent terminal collects and obtains viewing behavior data related to the user's viewing behavior in real time; Performing real-time content detection on the viewing behavior data collected in real time, extracting video metadata and picture features from the video content data in the viewing behavior data, and determining the preference of the current viewing user for different types of content; Performing real-time behavior detection on the viewing behavior data collected in real time, extracting features from the operation behavior data in the viewing behavior data, classifying and pattern-recognizing the extracted behavior features, and determining the operation behavior pattern of the current viewing user; At the same time, it conducts real-time environmental detection, monitors the viewing environment data of smart terminals, and assists in determining the identity information and viewing needs of the current viewing user; Generate a user profile for each time period of the current viewing user based on the determined preference, operation behavior pattern, identity information, and viewing needs of the current viewing user and the preset time period division rules; Based on the generated user portrait of the current viewing user in each time period, personalized recommended content is automatically provided to the current viewing user and recommended for display, as described above.

[0137] The step of generating a user portrait for each time period of the current viewing user based on the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user in combination with a preset time period division rule also includes: Establish a real-time feedback mechanism to continuously input new real-time detection data into the time-sharing portrait generation model, recalculate the current user's preference weight for content in the corresponding time period, and adjust the user portrait.

[0138] The step of the smart terminal acquiring viewing behavior data related to the user's viewing behavior in real time includes: The viewing behavior data related to the user's viewing behavior includes: viewing behavior data related to the user's active viewing behavior, and viewing behavior data related to randomly sending different content types to the user to test whether the user has a click preference for this type of content; Utilize multiple sensors and software monitoring mechanisms built into smart terminals to collect and obtain data related to user viewing behavior in real time; the viewing behavior data includes user operation data on the remote control, real-time feedback data of the terminal screen, voice command data of the user's interaction with the terminal, and application opening and usage data; The collected viewing behavior data is initially processed by data format conversion and / or data compression.

[0139] The step of performing real-time content detection on the viewing behavior data collected in real time, extracting video metadata and picture features from the video content data in the viewing behavior data, and determining the preference of the current viewing user for different types of content includes: Analyzing and preprocessing the viewing behavior data collected in real time; For video content data, extract the metadata and picture features of the video, and combine it with the user's operation behavior data on the video content, and use the content-based analysis algorithm to calculate the current user's real-time interest value for different types of content; Among them, the metadata includes data such as title, type and / or tag, the picture features include color distribution features and / or scene type features, and the operation behavior data includes viewing time data, whether to pause behavior data, and whether to fast-forward behavior data.

[0140] The steps of performing real-time behavior detection on the viewing behavior data collected in real time, extracting features from the operation behavior data in the viewing behavior data, classifying and pattern-recognizing the extracted behavior features, and determining the operation behavior pattern of the current viewing user include: Using a behavior analysis algorithm, performing sequence analysis and pattern recognition on the collected operation data of the viewing behavior data, analyzing the time sequence of user operations, and determining the rhythm and regularity of user operations; The machine learning model is used to classify and predict the user's operating behavior patterns, and determine the operating behavior pattern of the current viewing user.

[0141] The step of generating a user portrait for each time period of the current viewing user based on the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user and in combination with a preset time period division rule comprises: Obtain the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user, as well as the preset time period division rules; A combination of rule-based and machine learning methods is used. In terms of rule-based, initial portrait templates for different time periods are set; A machine learning algorithm is used to train and optimize the initial portrait template using historical data to obtain the probabilistic relationship between different detection results and user portrait features in each time period, so that the initial portrait template can be dynamically adjusted according to real-time detection data to obtain the user portrait of the current viewing user in each time period.

[0142] The step of automatically providing personalized recommended content to the current viewing user based on the generated user portrait of the current viewing user in each time period and displaying the recommended content includes: Provide personalized recommended content for the current viewing user based on the user portrait of each time period of the current viewing user; the recommended content includes video programs, applications, and advertising products, and the recommended content is dynamically adjusted according to the changes in the user's time-sharing portrait in different time periods; The recommended content is displayed in real time, as described above.

[0143] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), 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).

Claims

1. A family time-sharing portrait personality recommendation method based on real-time detection, characterized in that: include: The intelligent terminal collects and obtains viewing behavior data related to the user's viewing behavior in real time; Performing real-time content detection on the viewing behavior data collected in real time, extracting video metadata and picture features from the video content data in the viewing behavior data, and determining the preference of the current viewing user for different types of content; Performing real-time behavior detection on the viewing behavior data collected in real time, extracting features from the operation behavior data in the viewing behavior data, classifying and pattern-recognizing the extracted behavior features, and determining the operation behavior pattern of the current viewing user; At the same time, it conducts real-time environmental detection, monitors the viewing environment data of smart terminals, and assists in determining the identity information and viewing needs of the current viewing user; Generate a user profile for each time period of the current viewing user based on the determined preference, operation behavior pattern, identity information, and viewing needs of the current viewing user and the preset time period division rules; Based on the generated user portrait of the current viewing user in each time period, personalized recommended content is automatically provided to the current viewing user and displayed.

2. The method for recommending personality of family time-sharing portrait based on real-time detection according to claim 1 is characterized in that: The step of generating a user portrait for each time period of the current viewing user based on the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user and in combination with a preset time period division rule also includes: Establish a real-time feedback mechanism to continuously input new real-time detection data into the time-sharing portrait generation model, recalculate the current user's preference weight for content in the corresponding time period, and adjust the user portrait.

3. The method for recommending personality of family time-sharing portrait based on real-time detection according to claim 1 is characterized in that: The step of the smart terminal acquiring viewing behavior data related to the user's viewing behavior in real time comprises: The viewing behavior data related to the user's viewing behavior includes: viewing behavior data related to the user's active viewing behavior, and viewing behavior data related to randomly sending different content types to the user to test whether the user has a click preference for this type of content; Utilize multiple sensors and software monitoring mechanisms built into smart terminals to collect and obtain data related to user viewing behavior in real time; the viewing behavior data includes user operation data on the remote control, real-time feedback data of the terminal screen, voice command data of the user's interaction with the terminal, and application opening and usage data; The collected viewing behavior data is initially processed by data format conversion and / or data compression.

4. The method for recommending personality of family time-sharing portrait based on real-time detection according to claim 1 is characterized in that: The step of performing real-time content detection on the viewing behavior data collected in real time, extracting video metadata and picture features from the video content data in the viewing behavior data, and determining the preference tendency of the current viewing user for different types of content includes: Analyzing and preprocessing the viewing behavior data collected in real time; For video content data, extract the metadata and picture features of the video, and combine it with the user's operation behavior data on the video content, and use the content-based analysis algorithm to calculate the current user's real-time interest value for different types of content; Among them, the metadata includes data such as title, type and / or tag, the picture features include color distribution features and / or scene type features, and the operation behavior data includes viewing time data, whether to pause behavior data, and whether to fast-forward behavior data.

5. The method for recommending personality of family time-sharing portrait based on real-time detection according to claim 1 is characterized in that: The steps of performing real-time behavior detection on the viewing behavior data collected in real time, extracting features from the operation behavior data in the viewing behavior data, classifying and pattern-recognizing the extracted behavior features, and determining the operation behavior pattern of the current viewing user include: Using a behavior analysis algorithm, performing sequence analysis and pattern recognition on the collected operation data of the viewing behavior data, analyzing the time sequence of user operations, and determining the rhythm and regularity of user operations; The machine learning model is used to classify and predict the user's operating behavior patterns, and determine the operating behavior pattern of the current viewing user.

6. The method for recommending personality of family time-sharing portrait based on real-time detection according to claim 1, characterized in that: The step of generating a user portrait for each time period of the current viewing user according to the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user and in combination with a preset time period division rule comprises: Obtain the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user, as well as the preset time period division rules; A combination of rule-based and machine learning methods is used. In terms of rule-based, initial portrait templates for different time periods are set; A machine learning algorithm is used to train and optimize the initial portrait template using historical data to obtain the probabilistic relationship between different detection results and user portrait features in each time period, so that the initial portrait template can be dynamically adjusted according to real-time detection data to obtain the user portrait of the current viewing user in each time period.

7. The method for recommending personality of family time-sharing portrait based on real-time detection according to claim 1 is characterized in that: The step of automatically providing personalized recommended content to the current viewing user based on the generated user portrait of the current viewing user in each time period and displaying the recommended content includes: Provide personalized recommended content for the current viewing user based on the user portrait of each time period of the current viewing user; the recommended content includes video programs, applications, and advertising products, and the recommended content is dynamically adjusted according to the changes in the user's time-sharing portrait in different time periods; The recommended content is displayed in real time.

8. A family time-sharing portrait personality recommendation device based on real-time detection, characterized in that: The device comprises: A real-time data collection module, used to control the intelligent terminal to collect and obtain viewing behavior data related to the user's viewing behavior in real time; The content real-time detection submodule is used to perform real-time content detection on the viewing behavior data collected in real time, extract video metadata and picture features from the video content data in the viewing behavior data, and determine the preference of the current viewing user for different types of content; A behavior real-time detection submodule is used to perform behavior real-time detection on the viewing behavior data collected in real time, extract features of the operation behavior data in the viewing behavior data, classify and pattern recognize the extracted behavior features, and determine the operation behavior pattern of the current viewing user; The real-time environment detection submodule is used to simultaneously perform real-time environment detection, monitor the viewing environment data of the smart terminal, and assist in determining the identity information and viewing needs of the current viewing user; The family time-sharing portrait module is used to generate a user portrait for each time period of the current viewing user based on the determined preference tendency, operation behavior pattern, identity information and viewing needs of the current viewing user, combined with the preset time period division rules; The real-time portrait update submodule is used to establish a real-time feedback mechanism, continuously input new real-time detection data into the time-sharing portrait generation model, recalculate the current user's preference weight for content in the corresponding time period, and adjust the user portrait; A personalized recommendation engine module is used to automatically provide personalized recommended content to the current viewing user based on the generated user portraits of the current viewing user in each time period; The recommended content presentation module is used to display personalized recommended content.

9. An intelligent terminal, characterized in that: The device comprises a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, and the one or more programs include being used to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method as described in any one of claims 1 to 7.

Images