Video stream pushing method, electronic equipment, storage medium and product

Abstract

The invention provides a video stream pushing method, electronic equipment, a storage medium and a product. The method comprises the following steps: acquiring video streams of a plurality of visual angles corresponding to a target competition, competition data of a plurality of athletes and competition context information; obtaining competition watching behavior data of a target user from the client; determining whether a target event in which the target user is interested occurs in the target event within a preset time or not based on the video streams of the multiple perspectives, the competition participation data of the multiple athletes, the competition context information and the competition watching behavior data of the target user; and when it is determined that the target event occurs in the target event, pushing a video stream of a target visual angle associated with the target event in the video streams of the multiple visual angles to the client. According to the video stream pushing method and device, more accurate video stream pushing with a relatively high individuation degree can be realized.

Description

Technical Field

[0001] This application relates to the field of video technology, and in particular to a video streaming method, electronic device, storage medium, and product. Background Technology

[0002] With the development of internet technology and smart mobile devices, various internet products have brought convenience and enjoyment to people's work and life. Live video streaming has rapidly developed and been applied in people's daily work and life. Users can watch various live content, such as live sports events, through various live streaming platforms.

[0003] Currently, people are no longer satisfied with appreciating from a fixed perspective. In order to meet the diverse needs of live streaming, multiple camera positions can capture video streams from multiple angles for the same live streaming event. Therefore, the director's control of video streams from multiple angles will directly affect the user's live streaming viewing experience. Summary of the Invention

[0004] This application provides a video stream push method, electronic device, storage medium, and product, which can achieve more accurate and personalized video stream push.

[0005] The technical solution of this application embodiment is implemented as follows: This application provides a video stream push method, the method including: Acquire video streams from multiple perspectives, participation data of multiple athletes, and event context information corresponding to the target event; Obtain viewing behavior data from target users on the client side; Based on the video streams from the multiple perspectives, the participation data of the multiple athletes, the event context information, and the viewing behavior data of the target user, it is determined whether the target event of interest to the target user occurs within a preset time period. When it is determined that the target event has occurred in the target event, the video stream of the target perspective associated with the target event from the multiple video streams is pushed to the client.

[0006] This application provides an electronic device, the electronic device comprising: Memory is used to store executable instructions or computer programs. The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the video stream push method provided in the embodiments of this application.

[0007] This application provides a computer-readable storage medium storing a computer program or computer-executable instructions. When the computer program or computer-executable instructions are executed by a processor, they implement the video stream push method provided in this application.

[0008] This application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, they implement the video stream push method provided in this application.

[0009] The embodiments of this application have the following beneficial effects: The video stream push method, electronic device, storage medium, and product provided in this application, by integrating multi-view video streams corresponding to a target event, athlete participation data, event context information, and user viewing behavior data, can more accurately determine whether a target event of user interest has occurred within a preset time. When it is determined that a target event of user interest has occurred within the preset time, the video stream of the target viewpoint associated with the target event is selected from the multiple viewpoint video streams and pushed to the client, allowing the client to play the target viewpoint video stream so that the user can watch the event of user interest from the target viewpoint. This facilitates the accurate provision of personalized video streams that meet user preferences, enabling more accurate and highly personalized video stream pushes, thereby enhancing the user's live streaming viewing experience. Attached Figure Description

[0010] Figure 1 This application provides a schematic diagram of the system architecture for a video stream push method according to an embodiment of the present application. Figure 2 A flowchart illustrating a video stream push method provided in an embodiment of this application; Figure 3 A flowchart illustrating an application scenario of a video stream push method provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0013] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0014] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0015] Unless otherwise defined, all technical and scientific terms used in the embodiments of this application have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in the embodiments of this application is for the purpose of describing the embodiments of this application only and is not intended to limit this application.

[0016] In the implementation of this application, the collection and processing of relevant data should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0017] With the development of internet technology and smart mobile devices, various internet products have brought convenience and enjoyment to people's work and life. Live video streaming has rapidly developed and been applied in people's daily work and life. Users can watch various live content, such as live sports events, through various live streaming platforms.

[0018] Currently, people are no longer satisfied with viewing from a fixed perspective. Given the diverse needs of live streaming, multi-camera, multi-angle live streams can enrich the content and provide a more engaging viewing experience. Typically, for the same live event, there can be multiple video streams captured from different camera positions, offering various perspectives. Therefore, the director's control over these multi-angle video streams directly impacts the user's live streaming viewing experience.

[0019] In related technologies, the directing control of multi-view video streams mainly adopts a method of manual control or the use of preset simple directing rules (such as the trajectory of the ball and players in a stadium image) to uniformly push and switch between multiple clients. This results in multiple clients playing the same live broadcast at the same time, causing different users to see the same live broadcast at the same time. The directing control method is not flexible enough, lacks the ability to respond to individual user preferences, and is difficult to meet the diverse and personalized live broadcast needs of different users. This results in insufficient adaptation between the pushed video stream and the user's viewing needs, leading to a poor live broadcast viewing experience for users.

[0020] This application provides a video stream push method, electronic device, storage medium, and product. By integrating multimodal data from a sporting event (such as multi-view video streams, athlete participation data, event context information, and user viewing behavior data), it can more accurately determine whether an event of interest to the user has occurred or is about to occur. Furthermore, it can push a video stream from a target perspective related to the user's interest to the client at an appropriate time (such as when an event of interest occurs), allowing the user to view the event from that target perspective. This facilitates the accurate provision of personalized video streams that meet user preferences, effectively improving the accuracy of personalized video stream push and thus enhancing the user's live streaming viewing experience.

[0021] In some exemplary embodiments, the video streaming push methods provided in the various embodiments of this application are applicable to live sports events, such as live football matches, live basketball matches, live volleyball matches, and other event scenarios that can be broadcast from multiple perspectives.

[0022] The following describes an exemplary application of the video stream push method provided in the embodiments of this application. Figure 1 This is a schematic diagram of the system architecture for a video stream push method provided in an embodiment of this application. Figure 1 As shown, the system architecture may include: a server 10 and one or more terminal devices 20. Figure 1 (The example uses two terminal devices.) One or more terminal devices 20 communicate with the server 10 through network 30. Network 30 can be a wide area network (WAN), a local area network (LAN), or a combination of both.

[0023] Server 10 is used to acquire video streams from multiple perspectives for the same live event (such as a sports event), which may include: a video stream from the first perspective, a video stream from the second perspective, ..., a video stream from the Nth perspective, and to push a video stream from a certain perspective to terminal device 20 based on the video streams from these multiple perspectives, so that terminal device 20 can play the live video stream through the client. For example, N can be an integer greater than 2.

[0024] In some exemplary embodiments, the server 10 can be various electronic devices with certain computing capabilities, such as an independent physical server, a server cluster consisting of multiple physical servers, a distributed system consisting of multiple physical servers, a cloud server capable of providing cloud computing services, or a collaborative system consisting of edge nodes (such as NTP servers deployed at the edge) and cloud servers, etc. Here, the embodiments of this application do not limit this.

[0025] The terminal device 20 can be equipped with various clients for live video streaming, such as live streaming applications or live streaming mini-programs, allowing users to watch various live streaming content (such as live sports events) through the terminal device.

[0026] In some exemplary embodiments, the terminal device 20 can be various types of terminals, such as laptops, tablets, desktop computers, portable computers, smartphones, smart home appliances, vehicle terminals, or aircraft, etc. Here, the embodiments of this application do not limit this.

[0027] here, Figure 1 The example shown is merely a system architecture example of a video stream push method provided in the embodiments of this application, to help those skilled in the art understand the technical content of this application, but does not mean that the embodiments of this application cannot be used in other devices, systems, environments or scenarios.

[0028] Figure 2 This is a flowchart illustrating a video stream push method provided in an embodiment of this application. The following will be combined with... Figure 2 This will be explained. For example, this video stream push method can be used in... Figure 1 The server-side execution is shown. (Example) Figure 2 As shown, the video stream push method may include the following steps S201 to S204: Step S201: Obtain video streams from multiple perspectives, participation data of multiple athletes, and event context information corresponding to the target event; Step S202: Obtain viewing behavior data from target users on the client side; Step S203: Based on video streams from multiple perspectives, participation data of multiple athletes, event context information, and viewing behavior data of target users, determine whether the target event of interest to the target users has occurred within a preset time period. Step S204: When it is determined that a target event has occurred in the target event, push the video stream of the target viewpoint associated with the target event from multiple video streams to the client.

[0029] In some exemplary embodiments, the target event can be a sports event such as a football match or a basketball match.

[0030] In some exemplary embodiments, the video streams corresponding to the target event from multiple perspectives can be: multiple video streams captured by multiple camera positions at the event venue during the target event, with different shooting perspectives for each video stream. For example, at least one shooting perspective in the multiple video streams can be a global perspective, while some shooting perspectives in the multiple video streams can be the athlete's first-person perspective (referred to as the athlete's perspective).

[0031] Among them, each athlete's competition data refers to the data generated by the athlete during the competition at the target event venue, such as speed, heart rate, movements, emotions, etc., which can reflect the athlete's competition status.

[0032] Among them, event contextual information refers to relevant background and environmental data that can influence decision-making, analysis, and user experience in a target event (such as a football sports event), such as match schedule, event layout, and weather conditions.

[0033] In this context, "target users" refers to users who watch the target event through the client. For example, there can be one or more target users.

[0034] The target user's viewing behavior data refers to the behavioral data generated when the target user watches a sporting event (such as the current event or historical events). This data can reflect the target user's interests and preferences regarding the event, the events that occur during the event, and the participants. This data helps to more accurately predict when to switch perspectives and which perspective of the video stream will provide the best viewing experience, and helps to accurately provide users with personalized video streams that meet their preferences. For example, the target user's viewing behavior data can be collected through the client installed on the user's terminal device.

[0035] The target event can refer to a key match event occurring at the current moment in the target event or a key match event that will occur at the next moment. These key match events often last for a period of time. In some exemplary embodiments, taking a football match as an example, the target event can include, but is not limited to, one or more of the following events: shooting event, goal event, passing event, foul event, penalty event (such as the referee showing a yellow card or red card), substitution event, scoring event, mistake event, interception event, dribbling event, set piece event, penalty event, etc.

[0036] The preset time can refer to a preset time period, namely, from the first moment (i.e., the current moment, such as the 20th second of the second half of the match) to the second moment after the first moment (i.e., the 25th second of the second half of the match). Determining whether a target event of interest to the target user occurs in the target match within the preset time period can include any one or more of the following: determining whether a target event of interest to the target user is occurring in the target match at the first moment (i.e., the current moment), determining whether a target event of interest to the target user is about to occur in the target match between the first and second moments, and determining whether a target event of interest to the target user is about to occur in the target match at the second moment. Correspondingly, determining whether a target event occurs in the target match can include any one or more of the following: determining whether a target event of interest to the target user is occurring in the target match at the first moment (i.e., the current moment), determining whether a target event of interest to the target user is about to occur in the target match between the first and second moments, and determining whether a target event of interest to the target user is about to occur in the target match at the second moment. For example, the preset time can be 2 seconds, 3 seconds, 4 seconds, 5 seconds, or 6 seconds, etc., and this embodiment of the application does not limit this.

[0037] In some exemplary embodiments, the video stream of the target perspective associated with the target event can refer to a video stream that can better display the target event from a certain perspective or a video stream that can display the target event from a user-preferred perspective. For example, taking a shooting event in a football match as an example, the video stream of the target perspective can be a video stream from the first-person perspective of the player making the shooting action or a video stream from the first-person perspective of the goalkeeper, which is preferred by the user. In this way, after pushing the video stream of the target perspective to the client, the user can watch the entire shooting process from the first-person perspective of the player making the shooting action or from the first-person perspective of the preferred goalkeeper, thereby achieving an immersive user viewing experience.

[0038] Therefore, the video stream push method provided in this application, by integrating multi-view video streams corresponding to the target event, athlete participation data, event context information, and user viewing behavior data, can more accurately determine whether a target event of interest to the user has occurred (i.e., is currently occurring or about to occur) within a preset time. When it is determined that a target event of interest to the user has occurred within the preset time, the method selects the video stream of the target viewpoint associated with the target event from the multiple video streams and pushes it to the client, allowing the client to play the video stream of the target viewpoint so that the user can watch the event of interest from the target viewpoint. This facilitates the accurate provision of personalized video streams that meet user preferences, enabling more precise and personalized video stream pushes, effectively improving the accuracy of personalized video stream pushes, ensuring that the pushed video stream's perspective matches user preferences, preventing users from missing events of interest, reducing invalid or interfering perspective switching, and thus enhancing the user's live streaming viewing experience. Compared to related technologies that rely solely on manual or simple rules for video stream delivery, this application, in addition to multi-view video streams and athlete participation data, also incorporates user viewing behavior and event context information to achieve personalized video stream delivery. It can push video streams from target perspectives related to events that interest different users, ensuring that the live broadcast perspectives viewed by different users match their preferences. This enhances the user's live broadcast viewing experience, increases the personalization of viewing, and reduces invalid or disruptive switching.

[0039] In some exemplary embodiments, the video streams from multiple perspectives may include: a main-view video stream and multiple athlete-view video streams. The main-view video stream can refer to a global view captured by a data acquisition device positioned at a fixed camera location for the target event. In this main-view video stream, the athletes' movements are absolute, and the video captures a panoramic view of the competition, containing rich event information. This type of video stream typically displays overall or most of the information about the target event, such as the players' overall running trajectory and the ball's trajectory. The athlete-view video stream can refer to a first-person perspective view of a specific athlete captured by a data acquisition device positioned at a motion-capture camera location (such as a camera worn on the athlete's head). This type of video stream provides a more immersive viewing experience. In practical applications, there is a complementary relationship between the main-view and athlete-view video streams. The main-view video stream provides overall scene information, while the athlete-view video streams focus on the first-person perspective of a specific athlete. By dynamically switching between the main-view and athlete-view video streams based on user preferences or the current competition situation, a richer viewing experience can be achieved.

[0040] In some exemplary embodiments, each athlete's competition data may include one or more of the following: each athlete's physiological data, motion data, emotional data, and image data.

[0041] In some exemplary embodiments, the physiological data of each athlete may include, but is not limited to, biosignals such as heart rate, respiratory rate, body temperature, and blood oxygen saturation during the target competition. Analyzing these biosignals can reflect the athlete's physical condition and fatigue level. For example, wearable devices (such as physiological sensors) worn by athletes can be used to collect physiological data related to their exercise status, such as heart rate, respiratory rate, body temperature, and blood oxygen saturation, in real time.

[0042] In some exemplary embodiments, the motion data of each athlete may include, but is not limited to, data such as acceleration, position coordinates, limb movements, movement speed, movement trajectory, movement intensity, and running distance during the target event. Analyzing this motion data can assess the athlete's physical performance and movement efficiency. For example, this data can be acquired in real time through wearable devices worn by the athlete (such as motion sensors, motion capture sensors, accelerometers, gyroscopes, etc.), cameras at fixed positions, or video recording equipment worn by the athlete. For example, the motion sensor may be an inertial measurement unit (IMU). The IMU (Integrated Mutual Actuator) is small, lightweight, and can acquire a variety of motion data in real time and accurately.

[0043] In some exemplary embodiments, the emotional data of each athlete may include, but is not limited to, data on the athlete's excitement, agitation, frustration, tension, fatigue, etc., during the target competition. For example, emotional data can be obtained through facial expression recognition, voice analysis, and other means.

[0044] In some exemplary embodiments, the image data of each athlete may include, but is not limited to, facial image data, full-body image data, etc., of the athlete during the target event. For example, the image data can be acquired through a fixed-position camera, a camera device worn by the athlete, etc.

[0045] In some exemplary embodiments, the event context information may include one or more of the following: tactical setup, event stage, and event score. Tactical setup can refer to a team's formation, offensive and defensive strategies, such as whether to use a three-striker or two-striker configuration in attack. Event stage can be the current match time, remaining match time, first half, second half, extra time, penalty shootout, etc., with different stages having different rhythms and importance. The event score reflects the current outcome of the match and directly affects the user's emotional fluctuations and focus. Of course, the event context information may also include other background information that is important for determining the target event of interest to the user, such as team status and weather conditions. Thus, by combining event context information when determining the target event of interest to the user, more reasonable perspective switching decisions can be made at key moments (such as before a goal, after a shot, or in a penalty shootout), ensuring that the user does not miss any exciting moments while avoiding visual fatigue caused by frequent switching.

[0046] In some exemplary embodiments, step S201 may include: during the target event, acquiring video streams from multiple perspectives, participation data of multiple athletes, and event context information corresponding to the target event through a data acquisition device.

[0047] For example, data acquisition equipment can be deployed at the event venue and among athletes, and has high-frequency (e.g., no less than 30Hz) sampling accuracy and timestamp synchronization capabilities. For example, data acquisition equipment can consist of cameras at the event venue and wearable devices worn by athletes (such as cameras, IMU sensors, accelerometers, and gyroscopes).

[0048] In some exemplary embodiments, the video stream with multiple perspectives includes: a video stream from a main perspective and video streams from multiple athlete perspectives. Taking the event context information, including the event stage and the event score, as an example, step S201 may include the following steps: Step S2011: During the target event, the main view video stream is acquired in real time through cameras located at fixed positions on the field; Step S2012: During the target event, video streams from each athlete's perspective are acquired in real time through cameras worn on each athlete's head. Step S2013: During the target event, acquire the participation data of each athlete through wearable devices (such as IMU sensors, accelerometers, gyroscopes, etc.) worn by each athlete; Step S2014: During the target event, acquire event context information through data acquisition or data analysis equipment.

[0049] In some exemplary embodiments, the target user's viewing behavior data may include: current user behavior data of the target user watching the target event and historical user behavior data of the target user watching other events. This allows for the construction of a more user-preferred perspective switching timing judgment mechanism, more accurately determining whether key events occurring in a specific event align with the current user's preferences, achieving a more personalized viewing experience, and significantly enhancing immersion.

[0050] In some exemplary embodiments, the current user behavior data may include one or more of the following: physiological data generated by the target user watching the target event (such as user facial expressions, heart rate changes, volume, etc.), interaction data (such as bullet screen content, bullet screen density, total number of bullet screens, like frequency, pause / fast forward operation records, activity level, etc.), and emotional data (such as excitement, tension, agitation, etc.).

[0051] In some exemplary embodiments, historical user behavior data may include one or more of the following: viewing habits, viewing history, and followed athletes. In other exemplary embodiments, historical user behavior data may include one or more of the following: viewing habits, viewing history, followed athletes, and user type. For example, viewing habits may include preferred time periods, favorite genres, frequently followed teams, preferred winger performances in football matches, skipping defensive segments, etc. For example, viewing history may be a record of matches previously watched by the user. For example, user type may include analytical, entertainment-oriented, or celebrity-following users, etc.

[0052] In some exemplary embodiments, step S202 may include: obtaining the target user's viewing behavior data through a client in the terminal device (such as a mobile phone, tablet computer, etc.) used by the target user, and reporting the target user's viewing behavior data through the client in the terminal device.

[0053] In some exemplary embodiments, taking the target user's viewing behavior data as including: the target user's current user behavior data of watching the target event and the target user's historical user behavior data of watching other events, step S202 may include the following steps: Step S2021: While the target user is watching other events, the client collects and stores the target user's historical user behavior data; Step S022: During the target user's viewing of the target event, the client reports the target user's historical user behavior data; Step S2023: During the target user's viewing of the target event, the client collects and reports the target user's current user behavior data in real time.

[0054] Therefore, the video stream push method provided in this application, by integrating multimodal data corresponding to the target event, including multi-view video streams (such as a main-view video stream and multiple athlete-view video streams), athlete participation data (such as each athlete's physiological data, exercise data, and emotional data), event context information (such as tactical layout, event stage, and event score), and user viewing behavior data (such as current user behavior data and historical user behavior data), can more accurately determine whether a target event of interest to the user has occurred within a preset time. When it is determined that a target event of interest to the user has occurred within the preset time, the method selects the video stream of the target viewpoint associated with the target event from the multiple video streams and pushes it to the client, allowing the client to play the video stream of the target viewpoint, so that the target user can watch the event of interest from the target viewpoint. This facilitates the accurate provision of personalized video streams that meet user preferences, enabling more precise and personalized video stream pushes, effectively improving the accuracy of personalized video stream pushes, ensuring that the pushed video stream's perspective matches user preferences, preventing users from missing events of interest, reducing invalid or interfering perspective switching, and thus enhancing the user's live streaming viewing experience.

[0055] In some exemplary embodiments, step S203 may include steps S301 to S302: Step S301: Based on video streams from multiple perspectives, participation data of multiple athletes, and viewing behavior data of target users, detect whether a key event occurs in the target event within a preset time period using a preset key event detection model; Step S302: When a key event is detected in the target event within a preset time, based on the participation data of multiple athletes, event context information, and the viewing behavior data of the target user, determine whether the key event in the target event is a target event that the target user is interested in.

[0056] Among them, the preset key event detection model can comprehensively analyze multi-source heterogeneous data related to the target event (such as video streams from multiple perspectives, participation data of multiple athletes, and viewing behavior data of target users) to identify key events that may occur during the target event, such as shots, goals, red and yellow cards, and saves.

[0057] In some exemplary embodiments, the preset critical event detection model can be a deep learning-based multimodal fusion model, which refers to a neural network architecture that integrates multimodal data from the event (such as multi-view video streams, athlete participation data, event context information, and user viewing behavior data). For example, the preset critical event detection model can be implemented by combining a visual recognition model and a context-aware model.

[0058] In some exemplary embodiments, the pre-defined key event detection model can be implemented by combining 3D-Convolutional Neural Networks (3D-CNN) with a Transformer structure. Feature extraction and comprehensive analysis are performed using multi-view video streams (such as video frames), athlete participation data (such as athlete movement trajectories and the trajectory of a soccer ball), and event context information (such as the stage of the match). This, combined with sequence modeling using a Transformer structure, captures dynamic changes over time to identify potential key events in the event.

[0059] In other exemplary embodiments, the pre-defined critical event detection model can be implemented using a combination of Graph Neural Network (GNN) and Long Short-Term Memory (LSTM). The GNN extracts and comprehensively analyzes features from diverse information sources, including multi-view video streams (such as video frames), athlete participation data (such as athlete movement trajectories and the trajectory of a soccer ball), and event context information (such as the stage of the match). The LSTM learns from the dynamic changes in athlete participation behavior, event progress, and user viewing preferences over time. The outputs of the LSTM and GNN are then fused, classified, and predicted to identify potential critical events in the event.

[0060] In some exemplary embodiments, the output of a preset critical event detection model is a probability score for a critical event occurring in the target event within a preset time period (e.g., within the next n seconds, where n can be 2 seconds, 3 seconds, 4 seconds, 5 seconds, or 6 seconds, etc.). The probability score can be between 0 and 1; for example, the output of the preset critical event detection model might be a probability score of 0.85 for a shot occurring within the next 3 seconds. For instance, when the probability score of a critical event exceeds a preset score threshold (e.g., 0.75), it indicates that a critical event may occur in the target event within the preset time period. These critical events with probability scores exceeding the preset score threshold can then be considered as candidate critical events to further identify whether any of these candidate critical events are events of interest to the target user.

[0061] Thus, the video stream push method provided in this application analyzes video streams from multiple perspectives corresponding to a target event, participation data of multiple athletes, and viewing behavior data of target users through a preset key event detection model. This enables the detection of whether a key event occurs in the target event within a preset time. Subsequently, when a key event is detected in the target event within the preset time, the participation data of multiple athletes, event context information, and viewing behavior data of target users can be combined to further determine whether the key event in the target event is a target event of interest to the target user. In this way, when it is determined that a target event of interest to the target user has occurred in the target event, a video stream from the target perspective related to the target event can be pushed. In this way, by further determining whether the key events are of interest to the target users based on key event detection, it is possible to more accurately determine whether the target event of interest to the target users will occur within a preset time. This ensures the accuracy and matching degree of the identified target events of interest to the target users. As a result, it is beneficial to accurately provide users with video streams from personalized perspectives that meet their preferences. This enables more accurate and personalized video stream pushes, effectively improving the accuracy of personalized video stream pushes. It ensures that the perspective of the pushed video stream matches the user's preferences, preventing users from missing events of interest and reducing invalid or interfering perspective switching. Therefore, it can enhance the user's live streaming viewing experience.

[0062] In some exemplary embodiments, step S302: determining whether a key event occurring in the target event is a target event of interest to the target user, based on the participation data of multiple athletes, event context information, and the viewing behavior data of the target user, may include the following steps S3021 to S3022: Step S3021: Analyze the participation data of multiple athletes to determine their competition status; Step S3022: Based on the competition status of multiple athletes, event context information, and the viewing behavior data of the target user, determine whether the key events occurring in the target event are target events that the target user is interested in.

[0063] In some exemplary embodiments, taking the participation data of each athlete as including one or more of each athlete's physiological data, motion data, and image data as an example, step S3021 may include: determining at least one of each athlete's physiological state, motion state, and emotional state based on each athlete's participation data, wherein the emotional state is determined based on facial image data, the motion state is determined based on motion data, and the physiological state is determined based on physiological data; determining each athlete's competition state based on at least one of each athlete's physiological state, motion state, and emotional state, wherein the competition state includes: a first state for characterizing that the athlete is in a great competition state or a second state for characterizing that the athlete is not in a great competition state.

[0064] For example, when an athlete's competition data meets multiple indicators such as: the athlete's physiological state is strenuous exercise or stress, the athlete's exercise state is high-intensity exercise, and the athlete's emotional state is a high-emotional stage such as excitement and celebration, it can be determined that the athlete is in a great competition state.

[0065] In some other exemplary embodiments, step S302: determining whether a key event occurring in the target event is a target event of interest to the target user, based on the participation data of multiple athletes, event context information, and the viewing behavior data of the target user, may include the following steps S3023 to S3025: Step S3023: Based on whether the event context information meets the preset conditions, determine whether the key events occurring in the target event have switching value; Step S3024: When it is determined that the key event of the target event has switching value, the participation data of multiple athletes are analyzed, and athletes in excellent competition are identified from the multiple athletes as candidate athletes associated with the key event of the target event. Step S3025: Based on the target user's viewing behavior data and the candidate athletes associated with the key events of the target event, determine whether the key events of the target event are target events that the target user is interested in.

[0066] In some exemplary embodiments, the preset conditions can be pre-defined logical rules or threshold conditions used to determine the switching value of key events occurring in a target event during a certain period. For example, if the event context information indicates that the match is in the last five minutes and the scores of the two teams are relatively close, then under the event context information, any key event occurring in the target event could directly affect the match result. Therefore, the event context information can be determined to meet the preset conditions, and the value of switching perspectives under the event context information is high. Thus, the key events occurring in the target event can be determined to be important events with switching value.

[0067] In some exemplary embodiments, a "high-energy match state" refers to an athlete's high level of performance or emotional engagement at a particular moment. This is typically characterized by rapid movement, high energy expenditure, emotional excitement, and high participation in crucial moments, reflecting the relevance of the athlete to key events in the target competition. For example, taking a shooting event as a key event in the target competition, when an athlete breaks through multiple defenders in an attack and completes a shot, that athlete is in a "high-energy match state."

[0068] In some exemplary embodiments, by analyzing the participation data of all athletes, it is possible to identify which athletes played a key role in the critical events currently occurring in the target event, and mark these athletes in the most exciting moments of the competition as candidate athletes associated with the critical events of the target event. Subsequently, based on the user's viewing behavior data and the candidate athletes associated with the critical events of the target event, it is determined whether the target event of interest to the target user has occurred in the target event. This ensures that subsequent video stream recommendations are based not only on the importance of the critical events themselves, but also on the degree of matching between the athletes in the most exciting moments associated with the critical events and the user's viewing behavior data, thereby improving the accuracy of video stream recommendations.

[0069] In this way, by judging whether the key events of the target event have switching value through the event context information, and by combining the user's viewing behavior data and the candidate athletes associated with the key events of the target event to further verify whether the key events of the target event are the target events that the target users are interested in, the target events of the target event can be more accurately identified, and the target events can be ensured to conform to the event context information and fit the user's preferences. This can significantly enhance the user's immersion and provide a personalized viewing experience.

[0070] In some exemplary embodiments, step S3025: determining whether the key events of the target event are target events of interest to the target user based on the target user's viewing behavior data and the candidate athletes associated with the key events of the target event, may include steps S30251 to S30252: Step S30251: Based on the viewing behavior data and the candidate athletes associated with the key events of the target event, determine the target user's preference value for the key events of the target event; Step S30252: Based on the target user's preference value for key events of the target event, determine whether the key events of the target event are target events that the target user is interested in.

[0071] In some exemplary embodiments, step S30251 may include: scoring the key events of the target event based on the target user's viewing behavior data and candidate athletes associated with key events of the target event, to obtain preference values ​​for the key events of the target event. The preference values ​​represent the degree of preference the target user has for the key events of the target event. The preference values ​​reflect the user's level of attention to the key events; the higher the value, the more likely the user is to want to switch to watching from the relevant perspective of the key event.

[0072] In some exemplary embodiments, the scoring method may employ a pre-defined artificial intelligence (AI) model for scoring or a semantic classification-based quantitative evaluation mechanism (such as a weighted scoring algorithm).

[0073] For example, scoring key events in a target sporting event can include: using a pre-defined artificial intelligence (AI) model for scoring, based on the target user's viewing behavior data and candidate athletes associated with the key events in the target sporting event, to score the key events and obtain preference values ​​for those events. The AI ​​model comprehensively considers factors such as the user's historical preferences (e.g., liking a particular player or type of sport), real-time emotional feedback (e.g., increased heart rate, emotional expression), and the event's popularity (e.g., goals, saves, red cards), assigning a dynamic preference value to each key event in the target sporting event.

[0074] For example, scoring key events in a target event can include: weighting the key events in the target event based on the target user's viewing behavior data and the candidate athletes associated with the key events in the target event, and obtaining preference values ​​for the key events in the target event.

[0075] Weighted scoring refers to the process of assigning weight coefficients based on the importance of different semantic tags and calculating a comprehensive score.

[0076] For example, the preference value for different key events = Dribbling event (Boolean value) * Dribbling weight + Shooting event (Boolean value) * Shooting weight + Tackling event (Boolean value) * Tackling weight + Foul event (Boolean value) * Foul weight + Bullet comments (Boolean value) * Bullet comment weight * Number of bullet comments + User heart rate change (Boolean value) * Heart rate weight + User volume change (Boolean value) * Volume weight + ... + User-followed player (Boolean value) * Player weight + User preference type (Boolean value) * Type weight. Here, a Boolean value is a logical data type with only two values: true (usually represented by 1) and false (usually represented by 0). In the weighted scoring shown in this application embodiment, the Boolean value represents whether the content occurred or not. If the content occurred, the Boolean value is 1; if the content did not occur, the Boolean value is 0. For example, when a dribbling event occurs in the target event, the dribbling event (Boolean value) is 1; when no dribbling event occurs in the target event, the dribbling event (Boolean value) is 0.

[0077] For example, the preset preference threshold can be set based on information such as user type and user preferences. For instance, if a target user prefers to watch goal moments, a higher shooting weight can be set for shooting events.

[0078] In some exemplary embodiments, step S30252 may include: determining whether the preference value of a key event in the target event is greater than a preset preference threshold; when there are multiple key events in the target event whose preference value is greater than the preset preference threshold, taking all key events whose preference value is greater than the preset preference threshold as candidate key events; and determining the candidate key event with the highest preference value as the target event that the target user is interested in.

[0079] In some other exemplary embodiments, step S30252 may include: determining whether the preference value of a key event in the target event is greater than a preset preference threshold; when the number of key events in the target event whose preference value is greater than the preset preference threshold is one, the key event whose preference value is greater than the preset preference threshold is determined as the target event that the target user is interested in.

[0080] For example, the preset preference threshold can be set based on information such as user type and user preferences. For instance, when the target user type is tactical, a higher preset preference threshold can be set. When the target user type is a fan, a lower preset preference threshold can be set.

[0081] Thus, after scoring the key events of the target event, a preset preference threshold can be used to determine whether the key events reach a level of user interest. When the preference value of a key event is higher than the preset preference threshold, it can be determined that the key event is a target event of interest to the target user, so that the video stream of the target perspective associated with the target event from multiple perspectives can be pushed to the client. When the preference values ​​of all key events are not higher than the preset preference threshold, it indicates that the target user is not interested in the key event, and the key event can be ignored. In this case, there is no need to switch perspectives.

[0082] In some exemplary embodiments, taking a video stream with multiple perspectives, including a main perspective video stream and multiple athlete perspective video streams, as an example, after determining in step S3025 that a target event of interest to the target user occurs in the target event within a preset time, step S204: pushing the video stream of the target perspective associated with the target event from the multiple perspective video streams to the client, may include: Step S2041: Based on the candidate athletes associated with the key events of the target event, determine the video stream of the target perspective associated with the target event from the video streams of multiple perspectives; Step S2042: Send the video stream of the target viewpoint associated with the target event to the client.

[0083] For example, step S2042 may include the following steps: when the number of candidate athletes associated with the key event of the target event is one, directly determine the video stream of the athlete's perspective corresponding to the candidate athlete from the video streams of multiple perspectives as the video stream of the target perspective associated with the target event.

[0084] For example, step S2042 may include the following steps: when there are multiple candidate athletes associated with the key events of the target event, the target athlete preferred by the target user is determined from the multiple candidate athletes based on the target user's viewing behavior data or the target user's user type; the video stream from the perspective of the athlete corresponding to the target athlete preferred by the target user is determined as the video stream from the target perspective associated with the target event.

[0085] In some exemplary embodiments, taking a video stream with multiple perspectives including a main perspective video stream and multiple athlete perspective video streams as an example, step S204: pushing the target perspective video stream associated with the target event from the multiple perspective video streams to the client may include the following steps: taking the target athlete perspective video stream that meets the preset push conditions from the multiple athlete perspective video streams as the target perspective video stream, and pushing the target athlete perspective video stream to the client, so that the perspective of the video stream played by the client switches from the main perspective to the target athlete perspective.

[0086] For example, a video stream from the perspective of a target athlete that meets the preset push conditions can be: a video stream from the perspective of an athlete that corresponds to the target athlete preferred by the target user among multiple video streams; or a video stream from the perspective of an athlete that corresponds to an athlete in an exciting match among multiple video streams; or a video stream from the perspective of an athlete that can better showcase the target event among multiple video streams.

[0087] In some exemplary embodiments, taking a video stream with multiple perspectives including a main perspective video stream and multiple athlete perspective video streams as an example, step S204: pushing the target perspective video stream associated with the target event from the multiple perspective video streams to the client may include the following steps S401 to S402: Step S401: Determine whether the viewpoint of the current video stream pushed to the client is the main viewpoint; Step S402: When it is determined that the perspective of the current video stream pushed to the client is the main perspective, the current video stream pushed to the client is switched from the main perspective video stream to the target perspective video stream associated with the target event among multiple athlete perspective video streams, so as to switch the perspective of the current video stream pushed to the client from the main perspective to the athlete perspective.

[0088] In some exemplary embodiments, after step S402 switches the current video stream pushed to the client from the main viewpoint video stream to the target viewpoint video stream associated with the target event, the video stream pushing method may further include: Step S501: Determine whether the target event has ended; Step S502: When the target event is determined to have ended, the current video stream pushed to the client will revert to the main view video stream. Alternatively, when the target event ends, the main view video stream and multiple athlete view video streams will be pushed to the client simultaneously for the user to choose from.

[0089] In other exemplary embodiments, after step S402 switches the current video stream pushed to the client from the main viewpoint video stream to the target viewpoint video stream associated with the target event, the video stream pushing method may further include: Step S503: Determine whether the target event has not ended; Step S504: When it is determined that the target event has not ended, the current video stream pushed to the client is locked as the video stream of the target viewpoint associated with the target event until the target event ends.

[0090] In some exemplary embodiments, prior to step S201, the video stream push method may further include the following steps: Step S601: Obtain the raw multimodal data related to the target event; Step S602: Preprocess the raw multimodal data related to the target event to obtain video streams from multiple perspectives, participation data of multiple athletes, event context information, and viewing behavior data of target users. The preprocessing may include one or more of the following: data synchronization processing and data security processing.

[0091] In some exemplary embodiments, step S602 may include: using data synchronization techniques (such as time synchronization, timestamp matching, etc.) to synchronize the raw multimodal data related to the target event. For example, after collecting athlete sensor data through wearable devices (such as IMU sensors, accelerometers, gyroscopes, etc.) worn by each athlete, the athlete sensor data is denoised using denoising algorithms (such as median filtering, Kalman filtering); based on the denoised athlete sensor data, temporal features such as heart rate variability (HRV), motion intensity, and abrupt acceleration are extracted; based on the temporal features, the denoised athlete sensor data is real-time timestamped using an edge-deployed Network Time Protocol (NTP) server to obtain data after time synchronization and denoising. In this way, the participation data of each athlete is obtained, ensuring data consistency and analysis accuracy. Here, a timestamp refers to the alignment mark of multimodal data on the timeline.

[0092] In some other exemplary embodiments, step S601 may include: using data security technologies (such as federated learning mechanisms) to perform data security processing on sensitive data involving athletes and users in the raw multimodal data related to the target event. For example, for sensitive data of users and athletes, through federated learning mechanisms, feature extraction and model training are completed only on local devices to ensure data security and privacy compliance.

[0093] Figure 3 This is a flowchart illustrating an application scenario of a video stream push method provided in an embodiment of this application. Figure 3 The system architecture of an exemplary application scenario to which the video stream processing method of the embodiments of this application can be applied is shown. For example... Figure 3As shown, the system architecture may include: a data acquisition module, an event recognition module, an athlete status recognition module, a context association module, a decision-making module, and a video output control module. For example, the modules in the system architecture can be deployed and collaborated on by edge devices, a cloud computing platform, and user terminal devices. The process of this video stream push method may include: Step 1: The data acquisition module collects the following multimodal data in real time during the target event: video streams from multiple perspectives corresponding to the target event, participation data of multiple athletes and event context information, as well as viewing behavior data of the target users.

[0094] The video streams from multiple perspectives include: a main-view video stream and video streams from multiple athlete perspectives. Each athlete's participation data includes one or more of the following: physiological data, motion data, emotional data, and image data. The event context information includes one or more of the following: tactical deployment, event stage, and event score. The target user's viewing behavior data includes: current user behavior data of the target user watching the target event and historical user behavior data of the target user watching other events. Current user behavior data includes one or more of the following: physiological data, interaction data, and emotional data generated by the target user while watching the target event. Historical user behavior data includes one or more of the following: viewing habits, viewing history, and followed athletes.

[0095] For example, data acquisition equipment is deployed at the event site, in athletes' wearable devices, and in user terminals.

[0096] For example, the data acquisition module collects raw multimodal data in real time during the target event; performs unified timestamp alignment and feature extraction processing with the cloud server through edge nodes; and uses a federated learning framework to localize sensitive data related to athletes and users in the raw multimodal data, thereby obtaining the following multimodal data: video streams from multiple perspectives corresponding to the target event, participation data of multiple athletes and event context information, and viewing behavior data of target users.

[0097] Step 2: The event recognition module, based on video streams from multiple perspectives, competition data of multiple athletes, and viewing behavior data of target users, obtains a probability score of key events occurring in the target event within a preset time period (e.g., within the next n seconds, where n can be 2 seconds, 3 seconds, 4 seconds, 5 seconds, or 6 seconds, etc.) through a preset key event detection model (e.g., a combination of 3D convolutional neural network and Transformer structure).

[0098] In some exemplary embodiments, the output of the preset critical event detection model is a probability score of a critical event occurring in the target event within a preset time period (e.g., within the next n seconds, where n can be 2 seconds, 3 seconds, 4 seconds, 5 seconds, or 6 seconds, etc.).

[0099] Step 3: The athlete status recognition module analyzes the competition data of multiple athletes to determine their competition status. Competition status includes: a first status indicating an athlete is in an excellent performance state, and a second status indicating an athlete is not in an excellent performance state.

[0100] Step 4: The context association module determines whether a key event has occurred in the target event within a preset time period based on the probability score of the key event; it determines whether the key event in the target event has switching value based on whether the event context information meets preset conditions; when it is determined that the key event in the target event has switching value, it selects athletes in an exciting match from among the multiple athletes based on their match status, as candidate athletes associated with the key event in the target event, and uses the video stream from the perspective of the candidate athletes that meets the preset push conditions as the video stream from the target perspective associated with the target event.

[0101] For example, a video stream from the perspective of a candidate athlete that meets the preset push conditions can be: a video stream from the perspective of an athlete that corresponds to the target athlete preferred by the target user among multiple video streams, or a video stream from the perspective of an athlete that corresponds to an athlete in an exciting match among multiple video streams.

[0102] Step 5: The decision-making module, based on the target user's viewing behavior data (such as the user's past viewing history, followed players, viewing preferences, real-time physiological data, and real-time emotional data) and candidate athletes associated with key events in the target event, assigns weighted scores to the key events in the target event to obtain preference values ​​for these events. Based on whether the preference value of the key event is greater than a preset preference threshold, it determines whether the key event is a target event of interest to the target user. For example, if there are multiple key events with preference values ​​greater than the preset preference threshold, all key events with preference values ​​greater than the preset preference threshold are considered candidate key events; the candidate key event with the highest preference value is identified as the target event of interest to the target user. Similarly, if there is only one key event with a preference value greater than the preset preference threshold, the key event with a preference value greater than the preset preference threshold is identified as the target event of interest to the target user. After identifying the target events of interest to the target user, a switching instruction is generated.

[0103] Step 6: After receiving the switching instruction, the video output control module responds to the switching instruction by pushing the video stream of the target view associated with the target event from multiple video streams to the client.

[0104] In some exemplary embodiments, when the perspective of the current video stream pushed to the client is the main perspective, the current video stream pushed to the client is switched from the main perspective video stream to the target perspective video stream associated with the target event among multiple athlete perspective video streams, so as to switch the perspective of the current video stream pushed to the client from the main perspective to the athlete perspective.

[0105] In some exemplary embodiments, when the client is in a Virtual Reality (VR) or immersive environment, the current video stream pushed to the client is switched from the main viewpoint video stream to the target viewpoint video stream associated with the target event from among multiple athlete viewpoint video streams, and the panoramic view and spatial sound effects are adjusted simultaneously. This can further enhance the user's viewing experience.

[0106] In some exemplary embodiments, when the target event is detected to be ongoing, the viewpoint of the current video stream pushed to the client is locked, and the current video stream pushed to the client is locked as a video stream of the target viewpoint associated with the target event until the target event ends.

[0107] In some exemplary embodiments, when the target event is detected to have ended, the current video stream pushed to the client is reverted to a main-view video stream; or, a main-view video stream and multiple athlete-view video streams are pushed to the client simultaneously for the user to choose from.

[0108] In some exemplary embodiments, when it is determined that no key event has occurred in the target event, or when it is determined that the key event that has occurred in the target event is not a target event that the target user is interested in, a video stream from the main perspective and multiple video streams from the perspectives of athletes can be pushed to the client simultaneously for the user to choose from.

[0109] In some exemplary embodiments, when a video stream from the main perspective and multiple video streams from the perspectives of athletes are pushed to the client simultaneously, the client can play the video stream from the perspective of athletes associated with the target event in a small window.

[0110] In some exemplary embodiments, preset content (such as advertising content) is dynamically embedded in the video stream of the target view associated with the target event in the video stream of multiple viewpoints to obtain a processed video stream; the processed video stream is then pushed to the client.

[0111] Thus, the embodiments of this application not only realize the intelligence and personalization of event broadcasting, but also effectively take into account real-time performance, user privacy and system security. It can accurately provide users with video streams from personalized perspectives that meet their preferences, and can achieve more accurate and highly personalized video stream pushes, reducing invalid or interfering perspective switching, thereby enhancing the user's live viewing experience.

[0112] The embodiments of this application have the following beneficial effects: By integrating multimodal data corresponding to the target event, including multi-view video streams (such as video streams from the main perspective and video streams from multiple athlete perspectives), athlete participation data (such as physiological data, sports data, and emotional data of each athlete), event context information (such as tactical layout, event stage, and event score), and user viewing behavior data (such as current user behavior data and historical user behavior data), it is possible to more accurately determine whether a target event of interest to the user has occurred within a preset time. When it is determined that a target event of interest to the user has occurred within the preset time, the video stream from the target perspective associated with the target event is selected from the multiple video streams and pushed to the client, allowing the client to play the video stream from the target perspective, so that the target user can watch the event of interest from the target perspective. This facilitates the accurate provision of personalized video streams that meet user preferences, enabling more accurate and highly personalized video stream pushes, effectively improving the accuracy of personalized video stream pushes, ensuring that the pushed video stream perspective matches user preferences, preventing users from missing events of interest, reducing invalid or interfering perspective switching, and thus enhancing the user's live streaming viewing experience.

[0113] Compared to related technologies that rely solely on manual or simple rules for video stream delivery, this application, in addition to multi-view video streams and athletes' participation data (such as athletes' exercise status, physiological status, emotional status, etc.), also introduces users' viewing behavior data (such as volume changes, bullet screen density, user facial expressions, heart rate changes, user emotions) and event context information (such as tactical background, competition stage, etc.) to achieve personalized video stream delivery for users. In determining the target events of interest to users during a predetermined timeframe in a specific event, a user profiling system was established based on explicit preferences (such as user registration information and favorite players) and implicit preferences (such as viewing jumps, gaze points, and emotional fluctuations). This enhances the understanding of video stream footage, athlete participation data, match context information, and viewing behavior data. By combining data such as athlete trajectory analysis, athlete physiological and emotional states, tactical background, match stages, user facial expressions, heart rate changes, and user emotions, the importance of events occurring in the event can be comprehensively judged. This allows for accurate prediction of upcoming events, enabling predictive switching rather than delayed responses. Furthermore, by collecting multi-source heterogeneous data and establishing a multimodal analysis model, combined with user behavior and emotional feedback, a personalized switching decision mechanism can be constructed. This allows for a more accurate perception of user preferences, pushing video streams that match user tastes and emotions, achieving a precise, smooth, and immersive viewing experience. It also balances the need to enhance audience excitement with avoiding frequent switching, ensuring the integrity of key event perspectives (such as presenting shooting and celebration events in one integrated manner).

[0114] Embodiments of this application provide an electronic device, which may include: Memory is used to store executable instructions or computer programs. The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the video stream push method provided in one or more embodiments of this application.

[0115] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. This electronic device can be applied to... Figure 2 In a corresponding embodiment, a video stream push method is provided. For example... Figure 4 As shown, the electronic device 400 may include a processor 401, a memory 402, and a bus system 403. The various components in the electronic device 400 are coupled together via the bus system 403. Wherein: Bus system 403 is used to realize the communication connection between processor 401 and memory 402; Memory 402 is used to store computer-executable instructions or computer programs; When processor 401 executes computer-executable instructions or computer programs stored in memory 402, it performs the following steps: Acquire video streams from multiple perspectives, participation data of multiple athletes, and event context information corresponding to the target event; Obtain viewing behavior data from target users on the client side; Based on video streams from multiple perspectives, participation data of multiple athletes, event context information, and viewing behavior data of target users, it is determined whether the target event of interest to the target users occurs within a preset time period. When a target event is determined to occur in a target event, the video stream from the target perspective associated with the target event is pushed to the client from multiple video streams.

[0116] In some exemplary embodiments, the bus system 403 may include, in addition to a data bus, a power bus, a control bus, and a status signal bus, etc. However, for the sake of clarity, in Figure 4 The general designated all buses as Bus System 403.

[0117] In some exemplary embodiments, the electronic device may be implemented as a laptop computer, tablet computer, desktop computer, game console, server, or cloud computing platform.

[0118] In some exemplary embodiments, the processor may be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., wherein the general-purpose processor may be a microprocessor or any conventional processor, etc.

[0119] This application provides a computer-readable storage medium storing computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the video stream push method provided in this application can be implemented. For example, ... Figure 2 The video stream push method is shown.

[0120] This application provides a computer program product, which includes a computer program or computer-executable instructions. When the computer-executable instructions or the computer program are executed by a processor, the video stream push method provided in this application can be implemented. For example, ... Figure 2 The illustrated video streaming push method. For example, the computer program or computer-executable instructions are stored in a computer-readable storage medium, the processor of the electronic device reads the computer program or computer-executable instructions from the computer-readable storage medium, and the processor executes the computer program or computer-executable instructions, causing the electronic device to perform the video streaming push method provided in the embodiments of this application.

[0121] In some exemplary embodiments, the aforementioned computer-readable storage medium / memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM), etc.; or it may be various terminals including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0122] In some embodiments, a computer program or computer-executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0123] As an example, a computer program or computer-executable instructions may, but not necessarily, correspond to a file in a file system. It may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0124] As an example, a computer program or computer-executable instructions may be deployed to execute on one electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected by a communication network.

[0125] It should be noted that the descriptions of the above device, storage medium, or product embodiments are similar to the descriptions of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the device, storage medium, or product embodiments of this application, those skilled in the art should refer to the descriptions of the method embodiments of this disclosure for understanding. Further details will not be repeated here.

[0126] The features disclosed in the several methods, devices, storage media, or product embodiments provided in this application can be arbitrarily combined to obtain new method embodiments, devices, storage media, or product embodiments without conflict. The above descriptions are merely embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A video stream push method, characterized in that, The method includes: Acquire video streams from multiple perspectives, participation data of multiple athletes, and event context information corresponding to the target event; Obtain viewing behavior data from target users on the client side; Based on the video streams from the multiple perspectives, the participation data of the multiple athletes, the event context information, and the viewing behavior data of the target user, it is determined whether the target event of interest to the target user occurs within a preset time period. When it is determined that the target event has occurred in the target event, the video stream of the target perspective associated with the target event from the multiple video streams is pushed to the client.

2. The method according to claim 1, characterized in that, The video streams from multiple perspectives include: a video stream from the main perspective and video streams from multiple athlete perspectives; The step of pushing the video stream of the target perspective associated with the target event from the video streams of the multiple perspectives to the client includes: taking the video stream of the target athlete's perspective that meets the preset push conditions from the video streams of the multiple athlete's perspectives as the video stream of the target perspective, and pushing the video stream of the target athlete's perspective to the client, so that the perspective of the video stream played by the client is switched from the main perspective to the perspective of the target athlete.

3. The method according to claim 1, characterized in that, The video streams from multiple perspectives include: a video stream from the main perspective and video streams from multiple athlete perspectives; And / or, each athlete's competition data includes one or more of the following: each athlete's physiological data, athletic data, emotional data, and image data; And / or, the event context information includes one or more of the following: tactical setup, event stage, and event score; And / or, the target user's viewing behavior data includes: the target user's current user behavior data of watching the target event and the target user's historical user behavior data of watching other events. The current user behavior data includes one or more of the physiological data, interaction data, and emotional data generated by the target user while watching the target event. The historical user behavior data includes one or more of the following: viewing habits, viewing history, and followed athletes.

4. The method according to any one of claims 1 to 3, characterized in that, The method of determining whether a target event of interest to the target user occurs within a preset time period based on the video streams from the multiple perspectives, the participation data of the multiple athletes, the event context information, and the viewing behavior data of the target user includes: Based on the video streams from the multiple perspectives, the participation data of the multiple athletes, and the viewing behavior data of the target users, a preset key event detection model is used to detect whether a key event occurs in the target event within a preset time. When the key event is detected in the target event, based on the participation data of the multiple athletes, the event context information, and the viewing behavior data of the target user, it is determined whether the key event in the target event is a target event that the target user is interested in.

5. The method according to claim 4, characterized in that, The process of determining whether a key event in the target event is of interest to the target user, based on the participation data of the multiple athletes, the event context information, and the viewing behavior data of the target user, includes: Analyze the participation data of the multiple athletes to determine their competition status; Based on the competition status of the multiple athletes, the event context information, and the viewing behavior data of the target user, it is determined whether the key events occurring in the target event are target events that the target user is interested in.

6. The method according to claim 4, characterized in that, The process of determining whether a key event in the target event is of interest to the target user, based on the participation data of the multiple athletes, the event context information, and the viewing behavior data of the target user, includes: Based on whether the event context information meets preset conditions, it is determined whether the key events occurring in the target event have switching value; When it is determined that a key event in the target event has switching value, the participation data of the multiple athletes are analyzed, and athletes in excellent competitive condition are identified from the multiple athletes as candidate athletes associated with the key event in the target event. Based on the target user's viewing behavior data and the candidate athletes associated with key events in the target event, it is determined whether the key events in the target event are target events that the target user is interested in.

7. The method according to claim 6, characterized in that, The process of determining whether a key event in a target event is of interest to the target user, based on the target user's viewing behavior data and candidate athletes associated with key events in the target event, includes: Based on the target user's viewing behavior data and the candidate athletes associated with key events of the target event, determine the target user's preference value for the key events of the target event; Based on the target user's preference values ​​for key events occurring in the target event, determine whether the key events occurring in the target event are target events that the target user is interested in.

8. An electronic device, characterized in that, The electronic device includes: Memory is used to store executable instructions or computer programs. A processor, when executing computer-executable instructions or computer programs stored in the memory, implements the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program or computer-executable instructions, characterized in that, When the computer program or computer-executable instructions are executed by a processor, the method as described in any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program or computer-executable instructions, characterized in that, When the computer program or computer-executable instructions are executed by a processor, the method as described in any one of claims 1 to 7 is implemented.

Images