Motion action evaluation method, device and system, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a motion action evaluation method, device and system, electronic equipment and a storage medium. The motion action evaluation method comprises the steps of obtaining motion data of a target object in a target space; extracting motion feature data of the motion data; and according to the motion feature data, performing motion action evaluation to obtain a motion action evaluation result. By comprehensively and accurately acquiring and analyzing the exercise data, the exercise motion quality of multiple athletes can be evaluated in all directions and in real time, and the accuracy and comprehensiveness of exercise motion evaluation are effectively improved.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and particularly to a method, device, system, electronic device and storage medium for evaluating motion actions. Background Art

[0002] In the field of sports training, especially in high-tech sports like tennis, accurately evaluating and analyzing athletes' technical actions is crucial for improving training effects and competitive levels. Traditional action evaluation methods mainly rely on coaches' experience and subjective judgments. However, due to different possible judgment criteria among different coaches, the accuracy and consistency of evaluation results cannot be guaranteed. Secondly, coaches need to pay attention to the performances of multiple athletes simultaneously during training, making it difficult to analyze the hitting effects of each athlete in real time and comprehensively, and it is easy to miss some detailed problems. In addition, the manual evaluation method cannot provide quantitative data support, making it difficult to accurately track and analyze the training effects of athletes.

[0003] The foregoing description is provided to give general background information and does not necessarily constitute prior art. Summary of the Invention

[0004] Embodiments of this application provide a method, device, system, electronic device and storage medium for evaluating motion actions, which can simultaneously evaluate the quality of motion actions of multiple athletes in an all-round and real-time manner, effectively improving the accuracy and comprehensiveness of motion action evaluation.

[0005] In a first aspect, embodiments of this application provide a method for evaluating motion actions, including:

[0006] Obtaining motion data of a target object in a target space;

[0007] Extracting motion feature data from the motion data;

[0008] Performing a motion action evaluation based on the motion feature data to obtain a motion action evaluation result.

[0009] In a second aspect, embodiments of this application provide a device for evaluating motion actions, including:

[0010] A data acquisition module, configured to obtain motion data of a target object in a target space;

[0011] A feature extraction module, configured to extract motion feature data from the motion data;

[0012] An action evaluation module, configured to perform a motion action evaluation based on the motion feature data to obtain a motion action evaluation result.

[0013] Optionally, in some embodiments of the present application, the target space includes at least two target objects;

[0014] The step of extracting the motion feature data of the motion data includes:

[0015] Extracting the motion feature data of the motion data of the at least two target objects.

[0016] Optionally, in some embodiments of the present application, the target object is a sports person and / or a sports device held by the sports person, and the motion feature data is the human key element data of the sports person and / or the device key element data of the sports device.

[0017] Optionally, in some embodiments of the present application, the human key element data of the sports person at least includes the human key element data corresponding to the sports person holding the sports device.

[0018] Optionally, in some embodiments of the present application, the human key element data includes one or more combinations of the following: human key point data, line data, area data; the device key element data includes one or more combinations of the following: device key point data, line data, area data.

[0019] Optionally, in some embodiments of the present application, the device key point data includes one or more combinations of the following: the highest end point data of the sports device, the lowest end point data of the sports device, the leftmost end point data of the sports device, the rightmost end point data of the sports device, and / or the geometric center of gravity point data.

[0020] Optionally, in some embodiments of the present application, the feature extraction module includes:

[0021] An identification unit for identifying the type of motion action of the target object;

[0022] An extraction unit for extracting the motion feature data of the motion data according to the type of motion action.

[0023] Optionally, in some embodiments of the present application, the action evaluation module includes:

[0024] A synchronization unit for synchronizing the human key element data and the device key element data in the motion feature data;

[0025] An evaluation unit for performing a motion action evaluation based on the synchronized human key element data and device key element data to obtain a motion action evaluation result.

[0026] Optionally, in some embodiments of the present application, the synchronization unit includes:

[0027] A timing modeling subunit, configured to perform timing modeling on the human key element data and the device key element data to capture the timing dependency relationship of the action;

[0028] A boundary positioning subunit, configured to perform action boundary positioning on the human key element data and the device key element data after timing modeling, and generate a timing proposal for the start and end moments of the action;

[0029] A synchronization alignment subunit, configured to perform synchronization alignment processing on the human key element data and the device key element data in the time dimension according to the timing proposal.

[0030] Optionally, in some embodiments of the present application, the evaluation unit includes:

[0031] A timing analysis subunit, configured to perform timing analysis based on the human key element data and the device key element data after synchronization processing to obtain a timing analysis result;

[0032] An association analysis subunit, configured to perform association analysis based on the human key element data and the device key element data after synchronization processing to obtain an association analysis result;

[0033] An action evaluation subunit, configured to perform motion action evaluation based on the timing analysis result and the association analysis result to obtain a motion action evaluation result.

[0034] Optionally, in some embodiments of the present application, the evaluation unit is configured to:

[0035] Extract the shared features of the human key element data and the device key element data, and respectively generate human key element timing features and device key element timing features;

[0036] According to the human key element timing features and the device key element timing features, analyze the relative position relationship and change trend of the human key element data and the device key element data in the time dimension to obtain a motion action evaluation result.

[0037] Optionally, in some embodiments of the present application, the association analysis subunit is configured to:

[0038] Calculate the correlation between the human key element data and the device key element data after synchronization processing at different time points to obtain the association analysis result.

[0039] Optionally, in some embodiments of the present application, the types of motion actions include: serving, forehand hitting, two-handed backhand hitting, one-handed backhand hitting, forehand volley, backhand volley, forehand slice, backhand slice, overhead hitting.

[0040] Optionally, in some embodiments of the present application, the device further includes:

[0041] A personalized recommendation module, configured to generate personalized training recommendations based on the evaluation result of the motion actions.

[0042] Optionally, in some embodiments of the present application, the personalized recommendation module includes:

[0043] An acquisition unit, configured to acquire the body shape feature data of the athlete and / or the device characteristic data of the sports equipment;

[0044] A trajectory unit, configured to generate the real-time motion trajectory of the athlete based on the body shape feature data and / or the device characteristic data;

[0045] A comparison unit, configured to compare the real-time motion trajectory with the standard motion trajectory of the corresponding level of the athlete to obtain a motion trajectory comparison result;

[0046] A generation unit, configured to generate personalized training recommendations for the athlete based on the evaluation result of the motion actions and the motion trajectory comparison result.

[0047] Optionally, in some embodiments of the present application, the device further includes:

[0048] A technical report module, configured to acquire the historical evaluation result of the athlete's motion actions and generate a corresponding training technical report based on the historical evaluation result of the motion actions;

[0049] A feedback module, configured to feedback at least one of the evaluation result of the motion actions, the personalized training recommendations, and the training technical report to the smart mobile terminal of the athlete in real time.

[0050] In a third aspect, an embodiment of the present application further provides a motion action evaluation system, including a data acquisition device, a tennis service device, and a display device;

[0051] The data acquisition device is configured to acquire the motion data of the target object in the target space and send it to the tennis service device;

[0052] The tennis service device is configured to extract the motion feature data of the motion data and perform motion action evaluation according to the motion feature data to obtain a motion action evaluation result;

[0053] The display device is configured to feedback the motion action evaluation result in real time.

[0054] Fourthly, an embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it performs the steps of the motion action evaluation method as described in the first aspect.

[0055] Fifthly, an embodiment of the present application further provides a readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the computer program of the motion action evaluation method as described in the first aspect.

[0056] The present application provides a motion action evaluation method, device, system, electronic device, and storage medium. The motion action evaluation method includes: obtaining motion data of a target object in a target space; extracting motion feature data from the motion data; and performing motion action evaluation based on the motion feature data to obtain a motion action evaluation result. In the motion action evaluation solution provided by the present application, by comprehensively and accurately capturing various motion data of an athlete during the motion process, a reliable data basis is provided for subsequent feature extraction and action evaluation; the motion feature data in the motion data is accurately extracted, and thus multi-dimensional motion action evaluation is performed based on the motion feature data, which can not only improve the accuracy and efficiency of motion action evaluation, but also enhance the pertinence and real-time nature of training, and subsequently provide comprehensive and personalized technical feedback and training suggestions for the athlete, reduce the dependence on professional technical personnel or hardware devices, improve the training efficiency of the athlete, and reduce labor costs and equipment costs at the same time. Description of the Drawings

[0057] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings based on these drawings without creative efforts.

[0058] Figure 1 It is an application environment diagram of the motion action evaluation method provided by the embodiment of the present application;

[0059] Figure 2 It is an application environment diagram of the motion action evaluation method provided by another embodiment of the present application;

[0060] Figure 3 It is a flowchart of the motion action evaluation method provided by the embodiment of the present application;

[0061] Figure 4 It is a schematic diagram of key points when the target object is performing golf training provided by the embodiment of the present application;

[0062] Figure 5It is a schematic diagram of key points when the target object provided by the embodiment of the present application conducts table tennis training;

[0063] Figure 6 It is a schematic diagram of key points when the target object provided by the embodiment of the present application conducts tennis training;

[0064] Figure 7 It is another flowchart of the motion action evaluation method provided by the embodiment of the present application;

[0065] Figure 8 It is a schematic diagram of the structure of the motion action evaluation device provided by the embodiment of the present application;

[0066] Figure 9 It is another schematic diagram of the structure of the motion action evaluation device provided by the embodiment of the present application;

[0067] Figure 10 It is a schematic diagram of the structure of the electronic device provided by the embodiment of the present application. Detailed implementation manners

[0068] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present application. On the contrary, they are only examples of systems and methods consistent with the examples detailed in the appended claims or some aspects of the present application.

[0069] It should be noted that in this article, descriptions such as "including", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or elements inherent to such a process, method, article or device. Without further limitations, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including that element. In addition, components, features, and elements with the same name in different embodiments of the present application may have the same meaning or different meanings, and their specific meanings need to be determined according to their explanations in this specific embodiment or further in combination with the context of this specific embodiment.

[0070] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0071] In the following description, suffixes such as "module", "component", or "unit" used to denote elements are only for the convenience of describing the present application and have no specific meaning in themselves. Therefore, "module", "component", or "unit" can be used interchangeably.

[0072] Currently, in ball game training, most traditional technical action evaluation methods rely on coaches' experience for subjective judgment and lack accurate objective data support. As a result, athletes cannot obtain sufficient and timely technical feedback during the learning and improvement process, thus affecting the training effect. Although some action analysis methods based on computer vision and sensor technology have been proposed currently, these methods usually only focus on a single object, such as the human body or the racket, and fail to comprehensively consider the synergistic effect between the human body and the racket. However, in some ball game technical actions, the coordination between the human body and the racket is crucial, and analyzing only one aspect cannot comprehensively evaluate the technical state and technical problems of athletes. In addition, traditional methods usually adopt fixed standard actions or instructions, ignoring the individual differences of each athlete and unable to provide personalized feedback according to the specific characteristics of the athlete.

[0073] To solve the above technical problems, the embodiments of the present application provide a method, device, system, electronic device, and storage medium for evaluating motion actions. By comprehensively and accurately acquiring and analyzing motion data, it is possible to simultaneously evaluate the quality of motion actions of multiple athletes in an all-round and real-time manner, effectively improving the accuracy and comprehensiveness of motion action evaluation.

[0074] The motion action evaluation method provided by the embodiments of the present application can be applied in an application environment such as Figure 1 where the computer device 110 communicates with the server 120 through the network 130. The computer device 110 is used to acquire the motion data of the target object in the target space; extract the motion feature data of the motion data; and perform motion action evaluation according to the motion feature data to obtain the motion action evaluation result. In the present application, by comprehensively and accurately acquiring and analyzing motion data, it is possible to simultaneously evaluate the quality of motion actions of multiple athletes in an all-round and real-time manner, effectively improving the accuracy and comprehensiveness of motion action evaluation. Among them, the computer device 110 can be, but is not limited to, various smart phones 110-1, tablet computers 110-2, and laptop computers 110-3. The present application will be described in detail through specific embodiments below.

[0075] The motion action evaluation method provided by the embodiments of the present application can be applied in an application environment such as Figure 2The shown motion action evaluation system. The motion action evaluation system includes a data acquisition device 140, a tennis service device 160, and a display device 170; wherein, the data acquisition device 140 obtains the motion data of a target object 150 in a target space through a network 130 and sends it to the tennis service device 160; the tennis service device 160 extracts the motion feature data of the motion data through the network 130, and performs a motion action evaluation based on the motion feature data to obtain a motion action evaluation result; the display device 170 obtains the motion action evaluation result from the tennis service device 160 in real time through the network 130 and gives feedback.

[0076] The data acquisition device 140 is configured to obtain the motion data of the target object 150 in the target space and send it to the tennis service device 160;

[0077] Specifically, the data acquisition device 140 is used to obtain the motion data of the target object in the target space, such as high-precision sensors, depth cameras, intelligent wearable devices, etc. Combining multiple sensors and devices, such as optical sensors, inertial sensors, depth cameras, etc., comprehensively captures the motion data. Perform preprocessing operations on the collected raw data, such as filtering, denoising, enhancement, etc., to improve the data quality. Adopt high-speed data transmission technologies, such as 5G, Wi-Fi 6, etc., to ensure the real-time and continuity of the data.

[0078] The tennis service device 160 is configured to extract the motion feature data of the motion data, and perform a motion action evaluation based on the motion feature data to obtain a motion action evaluation result;

[0079] Specifically, the tennis service device 160 is used to process the motion data and generate an evaluation result, such as a server, a high-performance computer, etc. Utilize a deep learning model (such as HRNet) to extract high-resolution features and improve the accuracy of key point detection. At the same time, process multiple tasks, such as human key point detection, device key point detection, action recognition, etc., share underlying features, and improve efficiency. Utilize a cloud computing platform for data processing to improve the computing power and processing speed. Perform partial data processing near the data acquisition device to reduce data transmission latency and improve real-time performance.

[0080] The display device 170 is configured to give real-time feedback on the motion action evaluation result;

[0081] Specifically, the display device 170 is used to give real-time feedback on the motion action evaluation result, enabling the user to intuitively understand their hitting performance. The display device can be a screen, a mobile terminal, or other visualization devices, and the feedback form can include a graphical interface, real-time scoring, voice prompts, etc.

[0082] In addition, the display device can incorporate AR / VR technology to provide athletes with an immersive training experience. For example, it can display the evaluation results of sports movements in real time through AR glasses, or simulate different hitting scenarios in a VR environment to help athletes adapt to various competition environments. Meanwhile, a personalized feedback interface can be generated based on the training goals and historical data of the athletes. For example, more detailed movement guidance can be provided for beginners, and more accurate technical analysis can be provided for advanced athletes.

[0083] In a specific embodiment, the sports movement evaluation system can be integrated with the Internet of Things platform to achieve seamless communication and data sharing between devices. For example, the device status can be monitored in real time through the IoT platform to ensure the stable operation of the system. The system is designed to adapt to various training scenarios, such as standard tennis courts, serving machine training grounds, AR / VR virtual training environments, etc. For example, for the serving machine training ground, the evaluation algorithms for ball speed and landing point are optimized.

[0084] Please refer to Figure 3 , Figure 3 which is a schematic flowchart of the sports movement evaluation method provided by an embodiment of the present application. In this embodiment, the sports movement evaluation method is mainly exemplified by being applied to a computer device. The sports movement evaluation method provided by an embodiment of the present application includes the following steps:

[0085] S1. Obtain the movement data of the target object in the target space;

[0086] Specifically, for step S1, it is mainly to obtain the movement data corresponding to several target objects in the current target space. The acquisition method of the movement data can be data collection through high-precision sensors, depth cameras or smart wearable devices, and specific limitations are not imposed here. After the movement data is collected, preprocessing operations such as filtering and denoising can also be performed on the collected original movement data to improve the quality of the movement data. Among them, the target space in this embodiment refers to the environment where the sports movement occurs, such as a standard tennis court, a training ground or a virtual training platform (such as an AR / VR environment). The target object can include a sports person and / or the sports equipment (such as a tennis racket) held by him / her.

[0087] In addition, this embodiment can also combine multiple sensors for the collection of movement data, such as optical sensors, inertial sensors, etc., to improve the accuracy and robustness of the movement data. Or, high-speed cameras and real-time data transmission technology are adopted to ensure the real-time and continuous nature of the data.

[0088] This embodiment can comprehensively capture various information during the movement process, providing a rich data basis for subsequent feature extraction and motion evaluation. Through high-precision sensors and data preprocessing technologies, it ensures the accuracy and reliability of the collected data; real-time data acquisition and transmission technologies ensure the timeliness of the data, which is suitable for subsequent real-time training feedback and evaluation

[0089] S2. Extract the motion feature data of the motion data;

[0090] Specifically, for step S2, it is mainly to extract the motion feature data from the motion data obtained in step S1. Among them, the motion feature data can specifically include human key element data and device key element data. The human key element data can be the joint positions such as shoulders, elbows, wrists, knees, and ankles, as well as the posture information of parts such as the head and torso. The device key element data can be the racket face angle, racket head position, racket handle position, etc.

[0091] In a specific embodiment, by using a deep learning model (such as HRNet) to extract the high-resolution features of the motion data, the accuracy of key point detection is improved. At the same time, the detection tasks of human key points and device key points are processed, sharing the underlying features to improve the subsequent data processing efficiency. In addition, by introducing an attention mechanism, the correlation features between the athlete's hand and the racket handle are enhanced to improve the detection accuracy.

[0092] This embodiment can accurately extract motion feature data through deep learning algorithms and multi-task learning, improving the accuracy of key point detection. The efficiency of feature extraction is improved by sharing the underlying features and the multi-task learning framework, reducing the computational complexity. The added attention mechanism can effectively enhance the model's attention to key regions and improve the robustness in complex backgrounds.

[0093] S3. Perform a motion action evaluation based on the motion feature data to obtain a motion action evaluation result;

[0094] Specifically, for step S3, based on the extracted motion feature data, the quality and effect of the athlete's current motion action are evaluated. Among them, the evaluation indicators of the motion action evaluation can specifically include multiple dimensions such as accuracy, coordination, fluency, and swing speed. After completing the motion action evaluation, a comprehensive score is generated to evaluate the athlete's performance in this technical action, and the progress of their technical level is tracked according to the score trend. By comprehensively evaluating the motion action from multiple dimensions, the comprehensiveness and accuracy of the evaluation results are ensured.

[0095] In addition, personalized evaluation results and training suggestions can be generated according to the individual characteristics of the athlete and the device characteristics, and real-time feedback and suggestions are provided during the athlete's training process to help the athlete promptly correct problems in the technical actions.

[0096] The multi-dimensional evaluation system provided in this embodiment can ensure the comprehensiveness of the evaluation results, covering all aspects of the motion actions. The subsequent generated personalized evaluation and training suggestions can improve the pertinence and effectiveness of the training. And through the real-time feedback mechanism, it helps athletes obtain improvement suggestions in a timely manner during the training process and enhance the training effect.

[0097] It can be seen that the motion action evaluation method provided in this embodiment, by comprehensively, accurately and efficiently acquiring and processing motion data, extracting motion feature data, and performing multi-dimensional motion action evaluation based on the motion feature data, can provide comprehensive and personalized technical feedback and training suggestions for athletes, not only improving the accuracy and efficiency of motion action evaluation, but also enhancing the pertinence and real-time nature of subsequent training, and contributing to improving the technical level and training effect of athletes.

[0098] Optionally, in some embodiments, the target space in this embodiment may specifically include at least two target objects;

[0099] Then step S2, "extracting the motion feature data of the motion data", may specifically include:

[0100] Extracting the motion feature data of the motion data of at least two target objects.

[0101] Specifically, the target space refers to the environment where the motion action occurs, such as a standard tennis court, a training ground or a virtual training platform (such as an AR / VR environment). There are at least two target objects under the target space, including the athlete and / or the sports equipment held by him / her. For example, in a tennis training scenario, the target space can be a standard tennis court, and the target objects under the target space usually include the athlete and the tennis racket. In a tennis training scenario, the target space can be a standard tennis court, and the target objects include the athlete and the tennis racket held by him / her. The athlete is the main target object. The athlete performs various technical actions on the tennis court, such as serving, forehand hitting, backhand hitting, etc. The tennis racket is the secondary target object. The tennis racket is the equipment used by the athlete when performing technical actions, and its position and posture are crucial for the evaluation of technical actions.

[0102] In a specific embodiment, multiple target objects can be identified and distinguished through deep learning algorithms and computer vision technologies. For example, a convolutional neural network (CNN) is used to classify and locate different objects in the image. Track the motion trajectories of multiple target objects in a video sequence to ensure the continuity and consistency of the data. For example, a Kalman filter or a particle filter is used for target tracking. Analyze the interaction relationships between multiple target objects, such as the coordination between the athlete and the racket. For example, by calculating the relative position and motion trajectory between the athlete's hand and the racket, the accuracy of the swing action is evaluated.

[0103] Then, the motion feature data in the motion data corresponding to the moving person and / or the sports equipment held by the person are extracted respectively. For example, the motion feature data of the athlete's motion data are extracted, and the motion feature data of the tennis racket's motion data are extracted.

[0104] In a specific embodiment, a deep learning model (such as HRNet) can be used to extract high-resolution features to improve the accuracy of key point detection. For example, HRNet maintains high-resolution features through parallel multi-resolution subnets and repeated information exchange, especially for key points of small targets such as rackets. The detection tasks of human key points and equipment key points are processed simultaneously, sharing underlying features to improve efficiency. For example, two task heads are added based on HRNet, one for human key point detection and the other for racket key point detection. An attention mechanism is introduced to enhance the associated features between the hand and the racket handle to improve the detection accuracy. For example, the detection of racket key points is adjusted according to the positions of human key points.

[0105] This embodiment can extract motion feature data by processing the motion data of multiple target objects, comprehensively and accurately evaluate the quality and effect of motion actions, ensure the accuracy and reliability of data through multi-target object recognition and tracking technology, provide more in-depth action evaluation and feedback by analyzing the interaction relationships between target objects, and accurately extract motion feature data through deep learning algorithms and multi-task learning to improve the accuracy of key point detection.

[0106] Optionally, in some embodiments, the target object in this embodiment can specifically be a moving person and / or the sports equipment held by the moving person, and the motion feature data in this embodiment can specifically be the human key element data of the moving person and / or the equipment key element data of the sports equipment.

[0107] In a specific embodiment, the target object in this embodiment is a sports person and / or the sports equipment held by the sports person. A sports person refers to an individual participating in sports, such as a tennis player. Sports equipment refers to the equipment used by the sports person during the sports process, such as a tennis racket, a badminton racket, etc. The sports characteristic data is the human key element data of the sports person and / or the equipment key element data of the sports equipment. Among them, the human key element data includes human key point data, human key line data, human key area data, etc. The human key point data is, for example, the joint positions such as the shoulders, elbows, wrists, knees, and ankles, as well as the posture information of parts such as the head and torso. The human key line data: such as the relative positions and angles between various parts of the body, such as the connection angle between the shoulder and the elbow. The human key area data: such as the head area, the torso area, etc. The equipment key element data: includes equipment key point data, equipment key line data, equipment key area data, etc. The equipment key point data is, for example, the racket face angle, the racket head position, the racket handle position of the racket. The equipment key line data is such as the relative positions and angles between various parts of the racket, such as the angle between the racket face and the racket handle. The equipment key area data is such as the racket face area, the racket handle area, etc. In this embodiment, by comprehensively collecting the sports data of the sports person and the sports equipment and extracting the human key element data and the equipment key element data, the quality and effect of the sports action can be comprehensively and accurately evaluated.

[0108] Optionally, in some embodiments, the human key element data of the sports person in this embodiment at least includes the human key element data corresponding to the sports person holding the sports equipment.

[0109] Specifically, the human key element data of the sports person at least includes the human key element data corresponding to the sports person holding the sports equipment. The sports person holding the sports equipment refers to the sports person who holds the sports equipment during the sports process, such as a tennis player holding a tennis racket. The human key element data refers to the specific human key element data related to holding the sports equipment, such as the position and posture information of key points such as the hand, wrist, and elbow.

[0110] In a specific embodiment, a deep learning model (such as HRNet) can be used to extract hand key points to improve the accuracy of hand pose estimation. For example, HRNet maintains high-resolution features through parallel multi-resolution subnets and repeated information exchange, which is important for hand key point detection, especially for small targets such as finger key points. By analyzing the relative positions and movement trajectories between the hand key points and the equipment key points, the stability of the handheld device is evaluated. For example, the distance and angle between the hand key point and the racket handle are calculated to evaluate the stability of the racket grip. During the training process, feedback on the hand pose and equipment usage is provided in real time to help the athlete correct mistakes in a timely manner. For example, through a graphical interface or voice prompt, the athlete is informed in real time whether the racket grip is correct and whether the swing action is standard.

[0111] In this embodiment, by specifically focusing on the specific human key element data related to the handheld device, the technical movements of athletes when using the device can be evaluated more precisely. It not only improves the accuracy and pertinence of the evaluation, but also enhances the ability of real-time feedback and personalized training suggestions, providing athletes with more in-depth technical feedback and training guidance.

[0112] Optionally, in some embodiments, the human key element data in this embodiment may specifically include one or more of the following combinations: human key point data, line data, and region data; the device key element data in this embodiment may specifically include one or more of the following combinations: device key point data, line data, and region data.

[0113] Specifically, the human key element data in this embodiment includes one or more of the following combinations: human key point data, human key line data, and human key region data. Among them, the human key point data refers to the position information of each key part of the human body, such as the joint positions (elbows, knees, etc.), which are extracted through a deep learning model (such as HRNet) and can accurately represent the posture and movements of the human body. The human key line data refers to the connection information between each key point of the human body, such as the connection angle between the shoulder and the elbow, which can be used to analyze the relative positions and movement relationships between different parts of the human body. The human key region data refers to the information of specific regions of the human body, such as the head region, torso region, etc., which can be used to analyze the movement states and posture changes of different regions of the human body.

[0114] The device key element data includes one or more of the following combinations: device key point data, device key line data, and device key region data. Among them, the device key point data refers to the position information of each key part of the sports device, such as the racket face angle, racket head position, racket handle position, etc., which are extracted through a deep learning model and can accurately represent the posture and position of the sports device. The device key line data refers to the connection information between each key point of the sports device, such as the connection angle between the racket face and the racket handle, which can be used to analyze the relative positions and movement relationships between different parts of the sports device. The device key region data refers to the information of specific regions of the sports device, such as the racket face region, racket handle region, etc., which can be used to analyze the movement states and posture changes of different regions of the sports device.

[0115] In this embodiment, by simultaneously and detailedly extracting the human key element data and the device key element data, including key point data, line data, and region data, the postures and movements of the sports personnel and the sports device can be comprehensively and accurately described, improving the accuracy and efficiency of the evaluation.

[0116] Optionally, in some embodiments, the device key point data may specifically include one or more of the following combinations: the highest endpoint data of the sports device, the lowest endpoint data of the sports device, the leftmost endpoint data of the sports device, the rightmost endpoint data of the sports device, and / or the geometric center of gravity point data.

[0117] Specifically, the device key point data includes one or more of the following combinations: the highest endpoint data of the sports device, the lowest endpoint data, the leftmost endpoint data, the rightmost endpoint data, and / or the geometric center of gravity point data. Among them, the highest endpoint data refers to the position information of the highest point of the sports device in the vertical direction, such as the position of the racket head. The lowest endpoint data refers to the position information of the lowest point of the sports device in the vertical direction, such as the position of the racket handle. The leftmost endpoint data refers to the position information of the leftmost point of the sports device in the horizontal direction, such as the left edge of the racket. The rightmost endpoint data refers to the position information of the rightmost point of the sports device in the horizontal direction, such as the right edge of the racket. The geometric center of gravity point data refers to the position information of the geometric center of gravity of the sports device, such as the center of gravity position of the racket. Detecting tasks of multiple key points are processed simultaneously, sharing underlying features to improve efficiency. For example, adding multiple task heads on the basis of HRNet, which are respectively used to detect the highest endpoint, the lowest endpoint, the leftmost endpoint, the rightmost endpoint, and the geometric center of gravity point. An attention mechanism is introduced to enhance the attention to key regions and improve the detection accuracy. For example, adjust the detection of other key points according to the positions of the device key points.

[0118] In this embodiment, by simultaneously extracting the highest endpoint, the lowest endpoint, the leftmost endpoint, the rightmost endpoint, and the geometric center of gravity point data, the posture and position of the sports device can be comprehensively described, providing a reliable data basis for subsequent action evaluation; sharing underlying features and the multi-task learning framework improves the efficiency of feature extraction and reduces the computational complexity.

[0119] Optionally, in some embodiments, step S2, "extracting the motion feature data of the motion data", may specifically include:

[0120] S21. Identifying the motion action type of the target object;

[0121] Specifically, the target object refers to the sports person and / or the sports device held by him / her. By identifying the motion action type of the target object, it includes but is not limited to the specific action types performed by the target object during the motion, such as serving, forehand hitting, backhand hitting, forehand volley, backhand volley, forehand slice, backhand slice, overhead hitting, etc.

[0122] In a specific embodiment, a deep learning model (such as a convolutional neural network CNN) is used for action recognition to improve the recognition accuracy. For example, a pre-trained CNN model (such as ResNet-50) can be used to extract features from the input image or video data, and then the specific action type can be recognized through a classification layer.

[0123] S22. Extract the motion feature data of the motion data according to the motion action type;

[0124] Specifically, the motion feature data of the motion data is extracted according to the motion action type. The motion feature data refers to the key information extracted from the motion data, including human key element data, device key element data, etc.

[0125] In a specific embodiment, specific motion feature data is extracted according to different action types. For example, for the serving action, the key point data of the head, shoulders, elbows, wrists, etc. is extracted; for the forehand hitting action, the key point data of the shoulders, elbows, wrists, knees, etc. is extracted. Multiple feature extraction tasks are processed simultaneously, sharing the underlying features to improve efficiency. For example, multiple task heads are added based on HRNet to detect the key point data of different action types respectively. An attention mechanism is introduced to enhance the attention to the key areas and improve the detection accuracy. For example, according to the action type, the weight of the attention mechanism is adjusted to make the model pay more attention to the specific areas related to the action type.

[0126] In this embodiment, by identifying the motion action type of the target object and extracting the corresponding motion feature data according to the action type, the quality and effect of the motion action can be evaluated more accurately. It not only improves the accuracy and efficiency of the evaluation, but also provides rich information for the subsequent action evaluation, providing more in-depth technical feedback and training suggestions for the athletes.

[0127] Optionally, in some embodiments, step S3 "Perform motion action evaluation according to the motion feature data to obtain a motion action evaluation result" may specifically include:

[0128] S31. Synchronously process the human key element data and device key element data in the motion feature data;

[0129] Specifically, the motion feature data refers to the key information extracted from the motion data, including human key point data, device key point data, etc. The human key element data refers to the position information of each key part of the human body, such as the joint positions (elbows, knees, etc.). The device key element data refers to the position information of each key part of the motion device, such as the racket face angle, racket head position, racket handle position, etc. Align the human key element data and device key element data in the time dimension to ensure that both are analyzed under the same time stamp.

[0130] In a specific embodiment, a temporal feature extraction module (such as PP-TSM) is used to capture the inter-frame relationship and improve the accuracy of action recognition. PP-TSM performs feature exchange between the front and back time steps after each convolutional layer through lightweight temporal displacement operations, enabling the features of the current time step to fuse the information of the previous and next moments and enhancing the temporal modeling ability. A boundary matching network (such as BMN) is used to accurately locate the action boundary and improve the detection accuracy. BMN accurately locates the start time and end time of the action through boundary matching and generates potential temporal proposals, including the start and end moments of the action. A data fusion technology is adopted to fuse the human key element data and the device key element data to improve the accuracy of synchronous processing. For example, a Kalman filter or a particle filter can be used for data fusion to ensure the continuity and consistency of the data.

[0131] S32. Perform a motion action assessment based on the synchronized human key element data and device key element data after synchronous processing to obtain a motion action assessment result;

[0132] Specifically, the synchronized human key element data and device key element data after synchronous processing refer to the human key element data and device key element data that are aligned in the time dimension. Based on the synchronized data, the quality and effect of the motion action are evaluated to generate a motion action assessment result, including indicators such as the accuracy, coordination, and fluency of the action.

[0133] In a specific embodiment, the motion action is comprehensively evaluated from multiple dimensions, including accuracy, coordination, fluency, swing speed, etc. For example, a deep learning model (such as HRNet) can be used to extract high-resolution features to improve the accuracy of key point detection. According to the individual characteristics of the athlete and the characteristics of the racket, personalized assessment results and training suggestions are generated. For example, a deep learning model can be used to generate a standard action trajectory suitable for the individual and compare it with the standard action of professional athletes.

[0134] In this embodiment, by synchronously processing the human key element data and device key element data in the motion feature data, through temporal feature extraction and the boundary matching network, the human key element data and device key element data can be accurately aligned, improving the accuracy of synchronous processing; and based on the synchronized data, the motion action is comprehensively evaluated from multiple dimensions, enabling a more accurate assessment of the quality and effect of the motion action.

[0135] Optionally, in some embodiments, step S31 "Synchronously process the human key element data and device key element data in the motion feature data" may specifically include:

[0136] S311. Perform temporal modeling on the key element data of the human body and the key element data of the device to capture the temporal dependencies of actions;

[0137] Specifically, by processing the motion data to capture the temporal dependencies of actions. For example, the temporal feature extraction module can perform temporal modeling on the key element data of the human body and the key element data of the device. PP-TSM (Temporal Shift Module) performs feature exchange between the front and back time steps after each convolutional layer through lightweight temporal displacement operations, enabling the features at the current time step to fuse the information of the previous and subsequent moments, enhancing the temporal modeling ability. The temporal feature extraction module refers to the module used to extract the time series features in the motion data, such as PP-TSM (Temporal Shift Module). By using multi-scale feature extraction technology, capture the action features at different time scales to improve the accuracy of temporal modeling. Introduce the attention mechanism to enhance the attention to key time points and improve the accuracy of temporal modeling.

[0138] In this step, through the temporal feature extraction module, the temporal dependencies of actions can be accurately captured, improving the accuracy of action recognition. Through lightweight temporal displacement operations, the computational complexity is reduced, and the efficiency of temporal modeling is improved. Through multi-scale feature extraction and the attention mechanism, the robustness of the model in complex backgrounds is improved.

[0139] S312. Locate the action boundaries of the key element data of the human body and the key element data of the device after temporal modeling, and generate temporal proposals for the start and end moments of the action;

[0140] Specifically, by locating the action boundaries of the key element data of the human body and the key element data of the device after temporal modeling, potential temporal proposals are generated. For example, through the boundary matching network, the start and end times of the action can be determined, and the proposals for the start and end moments of the action generated include: BMN accurately locates the start time and end time of the action through boundary matching, generating potential temporal proposals that contain the start and end moments of the action. Among them, the boundary matching network refers to the network used to accurately locate the action boundaries, such as BMN (Boundary-Matching Network). Simultaneously process multiple action boundary location tasks, share the underlying features, and improve the efficiency. Introduce the attention mechanism to enhance the attention to key time points and improve the accuracy of action boundary location.

[0141] In this embodiment, through the boundary matching network, the start and end times of the action can be accurately determined, improving the accuracy of action boundary location; through multi-task learning and the attention mechanism, the efficiency of action boundary location is improved, and the computational complexity is reduced; through the boundary matching network and the attention mechanism, the robustness of the model in complex backgrounds is improved.

[0142] S313. Synchronously align the human key element data and the device key element data in the time dimension according to the timing proposal;

[0143] Specifically, according to the timing proposal, the human key element data and the device key element data are aligned in the time dimension, including: adopting data fusion technology to fuse the human key element data and the device key element data to improve the accuracy of synchronous alignment. Using a Kalman filter for data fusion to ensure the continuity and consistency of the data. Using a particle filter for data fusion to improve the robustness in complex backgrounds.

[0144] In this step, through the timing proposal, the human key element data and the device key element data can be accurately aligned, improving the accuracy of synchronous alignment. Through data fusion technology, the efficiency of synchronous alignment is improved and the computational complexity is reduced. Through the Kalman filter and the particle filter, the robustness of the model in complex backgrounds is improved.

[0145] In this embodiment, by performing timing modeling, action boundary localization, and synchronous alignment processing in the time dimension on the human key element data and the device key element data in the motion feature data, the quality and effect of the motion action can be evaluated more accurately.

[0146] Optionally, in some embodiments, step S32 "Perform motion action evaluation based on the synchronously processed human key element data and device key element data to obtain a motion action evaluation result" may specifically include:

[0147] S321. Perform timing analysis based on the synchronously processed human key element data and device key element data to obtain a timing analysis result;

[0148] Specifically, for step S321, time series analysis is performed on the human key element data and device key element data aligned in the time dimension to capture the timing characteristics and change trends of the actions, and a time series analysis result is obtained, including the timing characteristics and change trends of the actions. For example, a unified backbone network (such as HRNet) is used to extract shared features to ensure the retention of high-resolution features, which is important for key point detection, especially for small targets such as the key points of a racket. Multiple task heads are added on the basis of the backbone network, which are respectively used to generate human key point heatmaps and device key point heatmaps. A timing feature extraction module (such as PP-TSM) is used to capture the inter-frame relationship and improve the accuracy of action recognition. PP-TSM performs feature exchange between the front and back time steps after each convolutional layer through lightweight timing displacement operations, enabling the features at the current time step to fuse the information of the previous and subsequent moments and enhancing the timing modeling ability. Through the unified backbone network and multi-task learning, timing features can be accurately extracted, and the accuracy of time series analysis can be improved. Through sharing underlying features and the multi-task learning framework, the efficiency of feature extraction is improved, and the computational complexity is reduced.

[0149] S322. Perform correlation analysis based on the human key element data and device key element data after synchronous processing to obtain a correlation analysis result;

[0150] Specifically, for step S322, correlation analysis is performed on the human key element data and device key element data aligned in the time dimension to capture the mutual relationship between the human key elements and the device key elements, and a correlation analysis result is obtained, including the correlation between the human key elements and the device key elements. In addition, an attention mechanism can be introduced to enhance the attention to key regions and improve the accuracy of correlation analysis. For example, the detection of device key points is adjusted according to the positions of human key points. Multiple task heads are added on the basis of the backbone network, which are respectively used to detect human key points and device key points, sharing underlying features to improve efficiency. Feature fusion technology is adopted to fuse the human key element data and device key element data to improve the accuracy of correlation analysis. For example, methods such as weighted summation and feature splicing can be used for feature fusion.

[0151] In this step, through the attention mechanism and feature fusion technology, the mutual relationship between the human key elements and the device key elements can be accurately captured, and the accuracy of correlation analysis can be improved. Through sharing underlying features and the multi-task learning framework, the efficiency of feature extraction is improved, and the computational complexity is reduced. Through the attention mechanism, the attention of the model to key regions is enhanced, and the robustness in complex backgrounds is improved.

[0152] S323. Perform motion action evaluation based on the time series analysis result and the correlation analysis result to obtain a motion action evaluation result;

[0153] Specifically, for step S323, based on the timing analysis result and the association analysis result, a comprehensive evaluation of the motion action is performed to generate a motion action evaluation result, including indicators such as the accuracy, coordination, and fluency of the action.

[0154] In a specific embodiment, a comprehensive evaluation of the motion action is carried out from multiple dimensions, including accuracy, coordination, fluency, swing speed, etc. For example, a deep learning model (such as HRNet) can be used to extract high-resolution features to improve the accuracy of key point detection. According to the individual characteristics of the athlete and the characteristics of the racket, personalized evaluation results and training suggestions are generated. For example, a deep learning model can be used to generate a standard action trajectory suitable for the individual and compare it with the standard actions of professional athletes. Through the multi-dimensional evaluation system, the quality and effect of the motion action can be comprehensively evaluated, providing rich evaluation results. Through personalized evaluation, personalized training suggestions can be provided according to the specific characteristics of the athlete, improving the pertinence and effectiveness of training.

[0155] In this embodiment, by performing timing analysis and association analysis on the synchronized human key element data and device key element data, the quality and effect of the motion action can be evaluated more accurately, providing deeper technical feedback and training suggestions for the athlete.

[0156] Optionally, in some embodiments, step S32, "performing a motion action evaluation based on the synchronized human key element data and device key element data to obtain a motion action evaluation result", may specifically include:

[0157] Extracting the shared features of the human key element data and the device key element data, and respectively generating the human key element timing features and the device key element timing features;

[0158] Specifically, the shared features can be extracted from the human key element data and the device key element data by using a unified backbone network (such as HRNet). The shared features can capture the common patterns and changing trends of the human body and the device during motion. And through a timing feature extraction module (such as PP-TSM), the shared features are transformed into timing features, respectively generating the human key element timing features and the device key element timing features. The timing features can reflect the motion changes of the human body and the device in the time dimension.

[0159] In addition, this embodiment can also adopt multi-scale feature extraction technology to capture motion features at different time scales, improving the richness and accuracy of the timing features; introduce an attention mechanism to enhance the attention to key time points and key regions, improving the expression ability of the timing features; use feature fusion technology (such as weighted summation, feature splicing, etc.) to fuse different-scale and different-type features to further enhance the expression ability of the timing features.

[0160] Through shared feature extraction and temporal feature generation, this step can accurately capture the movement changes of the human body and the device in the time dimension, improving the accuracy of movement action evaluation; the use of a unified backbone network and temporal feature extraction module reduces the computational complexity of feature extraction and improves the evaluation efficiency; the introduction of multi-scale feature extraction and attention mechanism enhances the robustness of the model to different movement patterns and complex backgrounds.

[0161] Based on the temporal features of human key elements and the temporal features of device key elements, analyze the relative position relationship and change trend of human key element data and device key element data in the time dimension to obtain the movement action evaluation result;

[0162] Specifically, by analyzing the temporal features of human key elements and the temporal features of device key elements, determine the relative position relationship between human key elements and device key elements in the time dimension. For example, analyze the relative position changes of hand key points and racket key points during the swing. By analyzing the change trend of temporal features, evaluate the coordination and consistency of the human body and the device during the movement. For example, evaluate the fluency and stability of the swing action. Integrate the analysis results of the relative position relationship and change trend to generate the movement action evaluation result, including indicators such as action accuracy, coordination, and fluency.

[0163] In addition, temporal feature comparison techniques (such as dynamic time warping, Euclidean distance, etc.) can be used to quantify the relative position relationship and change trend of the human body and the device in the time dimension. Or use deep learning models (such as LSTM, GRU, etc.) to model the temporal features to further improve the accuracy and robustness of the analysis. Finally, visualize the analysis results to help athletes and coaches intuitively understand the movement action evaluation results.

[0164] This step can comprehensively evaluate the quality and effect of movement actions by analyzing the relative position relationship and change trend, providing rich evaluation results; provide personalized evaluation results and training suggestions according to the specific characteristics and movement patterns of athletes, improving the pertinence and effectiveness of training; through real-time analysis and feedback mechanism, ensure that the evaluation results can be provided to athletes and coaches in a timely manner, realizing real-time training feedback and evaluation.

[0165] Optionally, in some embodiments, step S322, "Perform correlation analysis based on the synchronized human key element data and device key element data to obtain the correlation analysis result", may specifically include:

[0166] Calculate the correlation between the synchronized human key element data and device key element data at different time points to obtain the correlation analysis result.

[0167] Specifically, for step S322, calculate the correlation between the human key element data and the device key element data after synchronous processing at different time points, that is, calculate the mutual relationship between the human key element data and the device key element data, such as the relative position and movement trajectory of the hand and the racket. Obtain the association analysis result through calculating the correlation, including the mutual relationship between the human key element and the device key element.

[0168] In a specific embodiment, multiple methods can be used to calculate the correlation, such as Euclidean distance, cosine similarity, Pearson correlation coefficient, etc. For example, the Euclidean distance between the hand key points and the racket key points can be calculated to evaluate the relative position relationship between the two. Use a deep learning model (such as HRNet) to extract high-resolution features to improve the accuracy of key point detection. For example, HRNet maintains high-resolution features through parallel multi-resolution subnets and repeated information exchange, which is important for key point detection, especially for small targets such as the key points of the racket. Introduce an attention mechanism to enhance the attention to key regions and improve the accuracy of correlation calculation. For example, adjust the detection of device key points according to the positions of human key points to make the model pay more attention to the relative position of the hand and the racket.

[0169] Through the correlation calculation method and deep learning algorithm in this embodiment, the correlation between the human key element data and the device key element data can be accurately calculated, improving the accuracy of association analysis; through sharing underlying features and a multi-task learning framework, the efficiency of feature extraction is improved and the computational complexity is reduced; through the attention mechanism, the attention of the model to key regions is enhanced, improving the robustness in complex backgrounds.

[0170] Optionally, in some embodiments, the types of motion actions in this embodiment may specifically include: serving, forehand hitting, two-handed backhand hitting, one-handed backhand hitting, forehand volley, backhand volley, forehand slice, backhand slice, overhead smash.

[0171] Specifically, the types of motion actions include at least one of serving, forehand hitting, two-handed backhand hitting, one-handed backhand hitting, forehand volley, backhand volley, forehand slice, backhand slice, overhead smash.

[0172] Taking the tennis training scenario as an example, serving refers to a basic technical movement in a tennis match. The player throws and hits the ball to the opponent's court. Serving is a movement that requires high-precision coordination and involves a large amount of body movement. Synchronously collect the head key points, left shoulder key points, right shoulder key points, left elbow key area, right elbow key area, left hand wrist key points, right hand wrist key points, left hip key area, right hip key area, left knee key points, right knee key points, left foot area, right foot area, racket head point, racket face area and racket handle position, and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis serving technical movement.

[0173] Forehand hitting refers to the technical movement of the player hitting the ball with the same side hand as the racket-holding hand. Forehand hitting is one of the most common hitting methods in tennis. Synchronously collect the left shoulder key points, right shoulder key points, left elbow key area, right elbow key area, left hand wrist key points, right hand wrist key points, left hip key area, right hip key area, left knee area, right knee, left foot area, right foot area, racket head point and racket face area; and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis forehand technical movement.

[0174] Two-handed backhand hitting refers to the technical movement of the player holding the racket with both hands and hitting the ball in a backhand manner; backhand hitting requires strong body coordination, especially the difference between one-handed backhand and two-handed backhand. For two-handed backhand hitting, synchronously collect the left shoulder key points, right shoulder key points, left elbow key area, right elbow key area, left hand wrist key points, right hand wrist key points, left hip key area, right hip key area, left knee key points, right knee key points, left foot area, right foot area, racket head point and racket face area; and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis two-handed backhand hitting technical movement.

[0175] One-handed backhand hitting refers to the technical movement of the player holding the racket with one hand and hitting the ball in a backhand manner; backhand hitting requires strong body coordination, especially the difference between one-handed backhand and two-handed backhand. For one-handed backhand hitting, synchronously collect the left shoulder key points, right shoulder key points, left elbow key area, right elbow key area, left hand wrist key points, right hand wrist key points, left hip key area, right hip key area, left knee key points, right knee key points, left foot area, right foot area, racket head point and racket face area, and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis one-handed backhand hitting technical movement.

[0176] Forehand volley refers to the technical movement of a player quickly hitting the ball in a forehand manner at the net; forehand volley is a technique that requires extremely high reaction speed and position selection. Synchronously collect key areas of the head, key points of the left shoulder, key points of the right shoulder, key areas of the left elbow, key areas of the right elbow, key points of the left wrist, key points of the right wrist, key areas of the left hip, key areas of the right hip, key points of the left knee, key points of the right knee, left foot area, right foot area, racket head point, and racket face area, and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis volley technical movement.

[0177] Backhand volley refers to the technical movement of a player quickly hitting the ball in a backhand manner at the net; synchronously collect key points of the head, key points of the left shoulder, key points of the right shoulder, key areas of the left elbow, key areas of the right elbow, key points of the left wrist, key points of the right wrist, key areas of the left hip, key areas of the right hip, key points of the left knee, key points of the right knee, left foot area, right foot area, racket head point, and racket face area, and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis volley technical movement.

[0178] Forehand slice refers to the technical movement of a player hitting the ball in a forehand manner to make the ball generate backspin; synchronously collect key points of the head, key points of the left shoulder, key points of the right shoulder, key areas of the left elbow, key areas of the right elbow, key points of the left wrist, key points of the right wrist, key areas of the left hip, key areas of the right hip, key points of the left knee, key points of the right knee, left foot area, right foot area, racket head point, and racket face area, and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis backhand slice technical movement.

[0179] Backhand slice refers to the technical movement of a player hitting the ball in a backhand manner to make the ball generate backspin; synchronously collect key points of the head, key points of the left shoulder, key points of the right shoulder, key areas of the left elbow, key areas of the right elbow, key points of the left wrist, key points of the right wrist, key areas of the left hip, key areas of the right hip, key points of the left knee, key points of the right knee, left foot area, right foot area, racket head point, racket face area, racket head position, and racket handle position, and analyze the relative positions of the above key points within the same time sequence, as well as the correlation analysis between different time points and these related positions, to achieve the analysis of the tennis forehand slice technical movement.

[0180] High-pressure hitting refers to the technical movement of an athlete hitting the ball with great force at a high position; synchronously collect the key points of the head, the key points of the left shoulder, the key points of the right shoulder, the key areas of the left elbow, the key areas of the right elbow, the key points of the left wrist, the key points of the right wrist, the key areas of the left hip, the key areas of the right hip, the key points of the left knee, the key points of the right knee, the left foot area, the right foot area, the racket head point, the racket face area, the racket head position, and the racket handle position, and analyze the relative positions of the above key points within the same time sequence according to the time sequence, as well as the correlation analysis between different time points and these related positions, so as to realize the analysis of the high-pressure hitting technical movement of tennis.

[0181] In this embodiment, different data collection methods are adopted for different types of sports movements. For example, for the serving movement, data on the athlete's ball-throwing movement and hitting movement can be collected; for the forehand hitting movement, data on the athlete's racket-swinging movement and hitting point can be collected. For different types of sports movements, different key point data are extracted. For example, for the serving movement, key point data such as the head, shoulders, elbows, and wrists can be extracted; for the forehand hitting movement, key point data such as the shoulders, elbows, wrists, and knees can be extracted. For different types of sports movements, different feature data are extracted. For example, for the serving movement, feature data such as the ball-throwing trajectory and hitting speed can be extracted; for the forehand hitting movement, feature data such as the racket-swinging trajectory and hitting force can be extracted. For different types of sports movements, different evaluation indexes are adopted. For example, for the serving movement, the accuracy of ball-throwing and the force of hitting can be evaluated; for the forehand hitting movement, the accuracy of racket-swinging and the stability of hitting can be evaluated.

[0182] It can be seen that by covering a variety of sports movement types, this embodiment can comprehensively evaluate the technical level of athletes and provide comprehensive technical feedback. By adopting different data collection, key point extraction, feature extraction, and movement evaluation methods for different types of sports movements, the quality and effect of each movement can be evaluated more accurately; personalized training suggestions can be provided according to the specific characteristics and movement types of athletes, improving the pertinence and effectiveness of training; feedback and suggestions can be provided in real time during the training process to help athletes correct problems in technical movements in a timely manner and improve the training effect.

[0183] Exemplarily, such as Figure 4As shown, when the target object is performing golf training, the recognized human key points and hitting equipment key points include head key point a11, neck key point a12, left shoulder key point a13, right shoulder key point a14, left hip key point a15, grip end key point a16, left knee key point a17, right knee key point a18, left ankle key point a19, right ankle key point a20, hose key point a21, sweet spot (optimal hitting position key point) a22, right elbow key point a24, and right wrist key point a25. Among them, a23 is the key area of the golf club. The grip end key point a16, hose key point a21, and sweet spot (optimal hitting position key point) a22, that is, the target key points, can be recognized through the detected key area of the golf club.

[0184] Exemplarily, as Figure 5 shown, when the target object is performing table tennis training, the recognized human key points and hitting equipment key points include head key point b11, neck key point b12, left shoulder key point b13, right shoulder key point b14, left wrist key point b15, grip key point b16, racket left shoulder key point b17, racket right shoulder key point b18, racket head key point b19, left hip key point b20, right hip key point b21, left elbow key point b23, and right elbow key point b24. Among them, b22 is the key area of the table tennis racket. The grip key point b16, racket left shoulder key point b17, racket right shoulder key point b18, and racket head key point b19, that is, the target key points, can be recognized through the detected key area of the table tennis racket.

[0185] Exemplarily, as Figure 6 shown, when the target object is performing tennis training, the recognized human key points and hitting equipment key points include head key point c11, neck key point c12, left shoulder key point c13, right shoulder key point c14, left elbow key point c15, left wrist key point c16, left hip key point c17, right hip key point c18, right elbow key point c19, right wrist key point c20, racket handle top key point c21, racket head top key point c22, racket head bottom key point c24, right knee key point c25, left knee key point c26, left ankle key point c27, and right ankle key point c28. Among them, c23 is the key area of the badminton racket. The racket handle top key point c21, racket head top key point c22, and racket head bottom key point c24, that is, the target key points, can be recognized through the detected key area of the tennis racket.

[0186] Optionally, as Figure 7 shown, in some embodiments, after obtaining the sports action evaluation result, the sports action evaluation method in this embodiment may specifically further include:

[0187] S4. Generate personalized training suggestions based on the sports action evaluation result.

[0188] Specifically, for step S4, the results obtained by evaluating the athlete's movement actions include scores in multiple dimensions such as the accuracy, coordination, and fluency of the actions. Accuracy refers to the precision of evaluating the racket swing action, measuring whether the action conforms to the standard trajectory. Coordination refers to evaluating the coordination of various parts of the human body and analyzing the coordinated cooperation of each part of the body during the racket swing. Fluency refers to evaluating the smoothness of the action and analyzing whether the trainee can smoothly complete the transition from the preparatory action to the hitting action. According to the specific situation and evaluation results of the athlete, targeted training suggestions are generated to help the athlete improve technical actions and enhance the training effect. For example, the technical level of the trainee is evaluated based on the action scoring results. The technical level of the trainee is divided into multiple levels, such as beginner, intermediate, and advanced. According to the trainee's scoring feedback, the technical level evaluation is updated in real time to help the trainee clarify their progress space.

[0189] In a specific embodiment, by deeply analyzing the evaluation results, the advantages and disadvantages in the athlete's technical actions are identified. For example, by comparing with the standard data of professional athletes, the deviations of the athlete at certain key points are found. Machine learning or deep learning algorithms are used to automatically generate personalized training suggestions according to the evaluation results. For example, the decision tree algorithm is used to recommend a specific training plan based on the evaluation results. A user profile of the athlete is constructed, including information such as their physical characteristics, technical features, and training history, in order to generate training suggestions that are more tailored to individual needs. A real-time feedback mechanism is established to enable the athlete to obtain improvement suggestions immediately during the training process. For example, the athlete's actions are monitored in real time through intelligent wearable devices, and voice or vibration prompts are provided.

[0190] Through in-depth analysis of the evaluation results, this embodiment can provide targeted training suggestions for athletes, helping them focus on improving the deficiencies in their technical actions. Considering the individual differences of athletes, training suggestions that conform to their personal characteristics are generated, improving the adaptability and effectiveness of training. The real-time feedback mechanism enables the athlete to obtain improvement suggestions immediately during the training process, adjust actions in a timely manner, and enhance the training effect; using intelligent algorithms to automatically generate training suggestions reduces the time for manual analysis and plan formulation, improving the efficiency of training.

[0191] Optionally, in some embodiments, step S4, "generating personalized training suggestions based on the evaluation results of movement actions", may specifically include:

[0192] S41. Obtain the body shape characteristic data of the athlete and / or the device characteristic data of the sports equipment;

[0193] Specifically, the body shape characteristic data of the athlete, that is, the body characteristic information of the athlete, such as height, weight, arm length, leg length, etc., can be measured and collected through intelligent wearable devices or depth cameras. The device characteristic data of the sports equipment refers to the physical characteristic information of the sports equipment, such as the weight, length, balance point, racket face angle of the racket, etc., which can be obtained through device sensors or preset parameters.

[0194] In a specific embodiment, intelligent wearable devices (such as smart bracelets, smart watches, etc.) can be used to collect the body shape characteristic data of the athlete in real time to ensure the accuracy and timeliness of the data; a depth camera is used to capture the body posture and movements of the athlete, and the body shape characteristic data is extracted through computer vision technology; sensors (such as acceleration sensors, gyroscopes, etc.) are installed on the sports equipment to collect the motion state and characteristic data of the equipment in real time; the data from different devices and sensors are fused to improve the integrity and accuracy of the data.

[0195] In this embodiment, through intelligent wearable devices and depth cameras, the body shape characteristic data of the athlete can be accurately collected, improving the accuracy of the data; the data is collected and transmitted in real time to ensure the timeliness and continuity of the data, which is applicable to real-time training feedback and evaluation; the device characteristic data of the sports equipment is collected to provide more comprehensive information for subsequent personalized training suggestions.

[0196] S42. Generate the real-time motion trajectory of the athlete based on the body shape characteristic data and / or the device characteristic data;

[0197] Specifically, the characteristic data and / or the device characteristic data are used to generate the real-time motion trajectory of the athlete, and the real-time motion trajectory refers to the real-time position and motion trajectory of the athlete during the movement. This includes the motion trajectories of the human body key points and the device key points. The collected body shape characteristic data and device characteristic data are combined with the real-time motion data of the athlete to generate a high-precision real-time motion trajectory.

[0198] In addition, a deep learning model (such as HRNet) can be used to extract high-resolution features to improve the accuracy of key point detection. For example, HRNet maintains high-resolution features through parallel multi-resolution subnets and repeated information exchange, which is important for key point detection, especially for small targets such as the key points of the racket. The human body key point data and the device key point data are fused to improve the accuracy of the motion trajectory. For example, methods such as weighted summation and feature splicing can be used for feature fusion. Real-time data processing technology is adopted to ensure that the generation of the motion trajectory can keep up with the real-time actions of the athlete and provide instant feedback.

[0199] In this embodiment, through deep learning algorithms and multi-task learning, it is possible to accurately generate the real-time motion trajectory of the moving person, improving the accuracy of the trajectory; generating the motion trajectory in real time to ensure the timeliness and continuity of the data, which is applicable to real-time training feedback and evaluation.

[0200] S43. Compare the real-time motion trajectory with the standard motion trajectory corresponding to the level of the moving person to obtain the motion trajectory comparison result;

[0201] Specifically, preset the standard motion trajectory according to the level of the moving person (such as beginner, intermediate, advanced) and the sports event (such as tennis, badminton, etc.). Compare the generated real-time motion trajectory with the standard motion trajectory, analyze the differences between the two, and obtain the motion trajectory comparison result. The motion trajectory comparison result obtained through comparison includes information such as the deviation and error of the motion trajectory. For example, compare the motion trajectory from multiple dimensions, such as the accuracy, smoothness, speed, etc. of the trajectory. Use error calculation methods (such as Euclidean distance, dynamic time warping, etc.) to quantify the differences in the motion trajectory. Visualize the comparison result to help the athlete intuitively understand the differences between their own motion trajectory and the standard trajectory. Provide the comparison result in real time during the training process to help the athlete adjust their movements in a timely manner.

[0202] This step can accurately analyze the differences in the motion trajectory through multi-dimensional comparison and error calculation, improving the accuracy of the comparison result; through visualization technology, visually display the comparison result to help the athlete better understand their own movement problems; provide the comparison result in real time to ensure the timeliness and continuity of the data, which is applicable to real-time training feedback and evaluation.

[0203] S44. Generate personalized training suggestions for the moving person based on the motion action evaluation result and the motion trajectory comparison result;

[0204] Specifically, the result obtained by evaluating the athlete's motion actions, including scores in multiple dimensions such as the accuracy, coordination, and smoothness of the actions, and the result obtained by combining the comparison of the real-time motion trajectory with the standard motion trajectory, including information such as the deviation and error of the motion trajectory, are used to generate targeted training suggestions to help the athlete improve their technical actions and enhance the training effect.

[0205] In addition, it is also possible to conduct in-depth analysis of the evaluation result and the comparison result to identify the strengths and weaknesses in the athlete's technical actions; use machine learning or deep learning algorithms to automatically generate personalized training suggestions based on the evaluation result and the comparison result; construct a user profile of the athlete, including information such as their physical characteristics, technical features, and training history, in order to generate training suggestions that are more tailored to individual needs; establish a real-time feedback mechanism to enable the athlete to obtain improvement suggestions immediately during the training process.

[0206] By deeply analyzing the evaluation results and comparison results, this embodiment can provide targeted training suggestions for athletes, helping them focus on improving the deficiencies in technical movements; considering the individual differences of athletes, generating training suggestions that suit their personal characteristics, and improving the adaptability and effectiveness of training; providing training suggestions in real time to ensure the timeliness and continuity of data, which is applicable to real-time training feedback and evaluation; using intelligent algorithms to automatically generate training suggestions, reducing the time for manual analysis and plan formulation, and improving the efficiency of training.

[0207] Optionally, as Figure 7 shown, in some embodiments, the movement action evaluation method in this embodiment may specifically further include:

[0208] S5. Obtain the historical movement action evaluation results of the athlete, and generate a corresponding training technology report based on the historical movement action evaluation results;

[0209] Specifically, obtaining the historical movement action evaluation results of the athlete refers to obtaining the movement action evaluation data of the athlete in previous trainings or competitions, including indicators such as the accuracy, coordination, and fluency of multiple evaluations. The report generated according to the historical evaluation results summarizes the athlete's technical level, progress trend, existing problems, and improvement suggestions. It can also provide long-term technical tracking services, helping to identify the progress trend and providing continuous technical improvement suggestions by analyzing the athlete's historical data.

[0210] For example, use cloud storage or a local database to store historical evaluation results to ensure data security and accessibility, use data mining and machine learning technologies to analyze historical data, identify the technical characteristics and improvement space of athletes, develop an automated report generation tool, present the analysis results in the form of charts, text, etc., for easy understanding by coaches and athletes, and use visualization technologies to display the technical progress trend of athletes and the changes in evaluation indicators, such as line charts and bar charts.

[0211] This embodiment provides a comprehensive technical evaluation by integrating historical data, helping athletes and coaches understand the long-term training effects; generating customized reports based on individual historical data to ensure that the suggestions and feedback meet the actual needs of athletes; the automated report generation tool reduces the time for manual collation and analysis, improving work efficiency; providing clear improvement suggestions through data analysis, providing a scientific basis for the formulation of training plans.

[0212] S6. Real-time feedback at least one of the movement action evaluation results, personalized training suggestions, and training technology reports to the intelligent mobile terminal of the athlete.

[0213] Specifically, the latest sports action evaluation results, personalized training suggestions, and training technology reports are immediately pushed to the athletes. The intelligent mobile terminal refers to devices such as smartphones and tablets used by athletes, which receive feedback information through a dedicated application. For example, through a dedicated mobile application, it supports the display of evaluation results, the push of training suggestions, and the download of reports. Using the push notification service to ensure that the evaluation results and suggestions can be delivered to the athletes in a timely manner. Implement data synchronization between the evaluation system and the mobile terminal to ensure the consistency and real-time nature of information. Design an intuitive and easy-to-use user interface to facilitate athletes to view the evaluation results and training suggestions.

[0214] In this embodiment, through the real-time feedback mechanism, it is ensured that athletes can immediately obtain the latest evaluation results and suggestions and adjust their training plans in a timely manner; by receiving information through the intelligent mobile terminal, athletes can view the feedback anytime and anywhere without being restricted by the venue; the mobile application can provide interactive functions such as feedback comments and setting training goals to enhance the athletes' sense of participation; timely feedback and visible progress trends can motivate athletes to maintain their training motivation and improve the training effect.

[0215] It should be noted that this embodiment can adapt to different training scenarios, including but not limited to tennis training scenarios, badminton training scenarios, table tennis training scenarios, and golf training scenarios.

[0216] In summary, the sports action evaluation method provided in this embodiment obtains the motion data of the target object in the target space; extracts the motion feature data of the motion data; and performs sports action evaluation based on the motion feature data to obtain the sports action evaluation result. In the sports action evaluation solution provided in this embodiment, by comprehensively and accurately capturing various motion data of athletes during the movement process, it provides a reliable data basis for subsequent feature extraction and action evaluation; accurately extracts the motion feature data in the motion data, and thus performs multi-dimensional sports action evaluation based on the motion feature data, which can not only improve the accuracy and efficiency of sports action evaluation, but also enhance the pertinence and real-time nature of training, and subsequently provide comprehensive and personalized technical feedback and training suggestions for athletes, reduce the dependence on professional technical personnel or hardware devices, improve the training efficiency of athletes, and at the same time reduce the labor cost and equipment cost.

[0217] It should be understood that although Figure 3 and Figure 7 the steps in the flowcharts of Figure 3 and Figure 7At least a part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages do not necessarily need to be completed at the same moment but can be executed at different moments. The execution order of these sub-steps or stages does not necessarily need to be sequential but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0218] To facilitate better implementation of the motion action evaluation method of the embodiments of the present application, embodiments of the present invention further provide a motion action evaluation device based on the above motion action evaluation method. The meanings of the nouns are the same as those in the above motion action evaluation method, and specific implementation details can be referred to the descriptions in the method embodiments.

[0219] Please refer to Figure 8 , Figure 8 which is a schematic structural diagram of the motion action evaluation device provided by the embodiments of the present application. The motion action evaluation device may specifically include a data acquisition module 201, a feature extraction module 203, and an action evaluation module 204, which may be specifically as follows:

[0220] The data acquisition module 201 is configured to acquire motion data of a target object in a target space;

[0221] The feature extraction module 202 is configured to extract motion feature data of the motion data;

[0222] The action evaluation module 203 is configured to perform motion action evaluation based on the motion feature data to obtain a motion action evaluation result.

[0223] Optionally, in some embodiments, the target space includes at least two target objects;

[0224] The steps of extracting motion feature data of the motion data include:

[0225] extracting motion feature data of the motion data of at least two target objects.

[0226] Optionally, in some embodiments, the target object is a sports person and / or a sports device held by the sports person, and the motion feature data is human key element data of the sports person and / or device key element data of the sports device.

[0227] Optionally, in some embodiments, the human key element data of the sports person at least includes human key element data corresponding to the sports person holding the sports device.

[0228] Optionally, in some embodiments, the human key element data includes one or more of the following combinations: human key point data, line data, area data; the device key element data includes one or more of the following combinations: device key point data, line data, area data.

[0229] Optionally, in some embodiments, the device key point data includes one or more combinations of the following: the highest end point data of the motion device, the lowest end point data of the motion device, the leftmost end point data of the motion device, the rightmost end point data of the motion device, and / or the geometric center of gravity point data.

[0230] Optionally, in some embodiments, the feature extraction module 202 includes:

[0231] An identification unit for identifying the motion action type of the target object;

[0232] An extraction unit for extracting the motion feature data of the motion data according to the motion action type.

[0233] Optionally, in some embodiments, the action evaluation module 203 includes:

[0234] A synchronization unit for synchronizing the human key element data and the device key element data in the motion feature data;

[0235] An evaluation unit for performing a motion action evaluation based on the synchronized human key element data and device key element data to obtain a motion action evaluation result.

[0236] Optionally, in some embodiments, the synchronization unit includes:

[0237] A timing modeling subunit for performing timing modeling on the human key element data and the device key element data to capture the timing dependency relationship of the action;

[0238] A boundary localization subunit for performing action boundary localization on the human key element data and the device key element data after timing modeling to generate a timing proposal for the start and end times of the action;

[0239] A synchronization alignment subunit for performing synchronization alignment processing on the human key element data and the device key element data in the time dimension according to the timing proposal.

[0240] Optionally, in some embodiments, the evaluation unit includes:

[0241] A timing analysis subunit for performing timing analysis based on the synchronized human key element data and device key element data to obtain a timing analysis result;

[0242] An association analysis subunit for performing association analysis based on the synchronized human key element data and device key element data to obtain an association analysis result;

[0243] An action evaluation subunit, configured to perform a motion action evaluation based on the time series analysis result and the correlation analysis result to obtain a motion action evaluation result.

[0244] Optionally, in some embodiments, the evaluation unit is configured to:

[0245] Extract the shared features of the human key element data and the device key element data, and respectively generate a human key element time series feature and a device key element time series feature;

[0246] According to the human key element time series feature and the device key element time series feature, analyze the relative position relationship and change trend of the human key element data and the device key element data in the time dimension to obtain a motion action evaluation result.

[0247] Optionally, in some embodiments, the correlation analysis subunit is configured to:

[0248] Calculate the correlation between the synchronized human key element data and the device key element data at different time points to obtain a correlation analysis result.

[0249] Optionally, in some embodiments, the types of motion actions include: serving, forehand hitting, two-handed backhand hitting, one-handed backhand hitting, forehand volley, backhand volley, forehand slice, backhand slice, overhead smash.

[0250] Optionally, as Figure 9 shown, in some embodiments, the device further includes:

[0251] A personalized recommendation module 204, configured to generate personalized training recommendations based on the motion action evaluation result.

[0252] Optionally, in some embodiments, the personalized recommendation module includes:

[0253] An acquisition unit, configured to acquire the body shape feature data of the athlete and / or the device characteristic data of the sports equipment;

[0254] A trajectory unit, configured to generate a real-time motion trajectory of the athlete based on the body shape feature data and / or the device characteristic data;

[0255] A comparison unit, configured to compare the real-time motion trajectory with the standard motion trajectory of the corresponding level of the athlete to obtain a motion trajectory comparison result;

[0256] A generation unit, configured to generate personalized training recommendations for the athlete based on the motion action evaluation result and the motion trajectory comparison result.

[0257] Optionally, as Figure 9 shown, in some embodiments, the device further includes:

[0258] A technical report module 205 is configured to obtain historical motion action evaluation results of a moving person and generate a corresponding training technical report based on the historical motion action evaluation results;

[0259] A feedback module 206 is configured to feedback at least one of the motion action evaluation results, personalized training suggestions, and training technical reports to the intelligent mobile terminal of the moving person in real time.

[0260] For the specific limitations of the motion action evaluation device, reference can be made to the limitations of the motion action evaluation method in the foregoing text, which will not be elaborated here. Each module in the foregoing motion action evaluation device can be implemented in whole or in part by software, hardware, and their combination. The foregoing modules can be embedded in the processor in the computer device in hardware form or independent of the processor, or stored in the memory in the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to the foregoing modules.

[0261] The motion action evaluation device provided in this embodiment obtains motion data of a target object in a target space through a data acquisition module 201; extracts motion feature data of the motion data through a feature extraction module 202; and performs a motion action evaluation according to the motion feature data through an action evaluation module 203 to obtain a motion action evaluation result. This embodiment comprehensively and accurately captures various motion data of the athlete during the motion process, providing a reliable data basis for subsequent feature extraction and action evaluation; accurately extracts the motion feature data in the motion data, and thus performs a multi-dimensional motion action evaluation according to the motion feature data, which can not only improve the accuracy and efficiency of the motion action evaluation, but also enhance the pertinence and real-time nature of the training, so as to subsequently provide comprehensive and personalized technical feedback and training suggestions for the athlete, reduce the dependence on professional technical personnel or hardware devices, improve the training efficiency of the athlete, and reduce the labor cost and equipment cost at the same time.

[0262] In addition, an embodiment of the present application further provides an electronic device, as Figure 10 shown, which shows a schematic structural diagram of the electronic device involved in the embodiment of the present application. Specifically:

[0263] The electronic device may include a processor 301 with one or more processing cores, a memory 302 with one or more computer-readable storage media, a power supply 303, an input unit 304, and other components. Those skilled in the art can understand that Figure 10 the structural diagram of the electronic device shown in

[0264] The processor 301 is the control center of the electronic device, connecting various parts of the entire electronic device through various interfaces and circuits. By running or executing software programs and / or modules stored in the memory 302, and calling the data stored in the memory 302, it executes various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor. Among them, the application processor mainly processes the operating system, user interface, application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor may not be integrated into the processor 301 either.

[0265] The memory 302 can be used to store software programs and modules. The processor 301 executes various functional applications and motion action evaluation methods by running the software programs and modules stored in the memory 302. The memory 302 mainly includes a program storage area and a data storage area. Among them, the program storage area can store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.); the data storage area can store data created according to the use of the electronic device. In addition, the memory 302 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices. Correspondingly, the memory 302 may also include a memory controller to provide the processor 301 with access to the memory 302.

[0266] The electronic device also includes a power supply 303 that powers each component. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system. The power supply 303 may also include any components such as one or more DC or AC power supplies, a recharge system, a power failure detection circuit, a power converter or inverter, and a power status indicator.

[0267] The electronic device may also include an input unit 304, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

[0268] Although not shown, the electronic device may also include a display unit, etc., which will not be elaborated here. Specifically, in this embodiment, the processor 301 in the electronic device will load the executable files corresponding to the processes of one or more application programs into the memory 302 according to the following instructions, and the processor 301 will run the application programs stored in the memory 302 to achieve various functions as follows:

[0269] Obtain the motion data of the target object in the target space; extract the motion feature data of the motion data; perform a motion action evaluation based on the motion feature data to obtain a motion action evaluation result.

[0270] For the specific implementation of each of the above operations, reference may be made to the previous embodiments, which will not be elaborated here.

[0271] In the embodiments of the present application, by comprehensively and accurately capturing various motion data of athletes during the motion process, a reliable data basis is provided for subsequent feature extraction and action evaluation; the motion feature data in the motion data is accurately extracted, so as to perform multi-dimensional motion action evaluation based on the motion feature data, which can not only improve the accuracy and efficiency of the motion action evaluation, but also enhance the pertinence and real-time nature of training, and thus provide comprehensive and personalized technical feedback and training suggestions for athletes subsequently, reduce the dependence on professional technical personnel or hardware devices, improve the training efficiency of athletes, and reduce the labor cost and equipment cost at the same time.

[0272] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by controlling related hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0273] Therefore, the embodiments of the present application provide a storage medium, in which multiple instructions are stored, and the instructions can be loaded by a processor to execute the steps in any of the motion action evaluations provided by the embodiments of the present application. For example, the instructions can execute the following steps:

[0274] Obtain the motion data of the target object in the target space; extract the motion feature data of the motion data; perform a motion action evaluation based on the motion feature data to obtain a motion action evaluation result.

[0275] For the specific implementation of each of the above operations, reference may be made to the previous embodiments, which will not be elaborated here.

[0276] Among them, the storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc.

[0277] Since the instructions stored in the storage medium can execute the steps in any of the motion action evaluation methods provided by the embodiments of the present application, the beneficial effects that can be achieved by any of the motion action evaluation methods provided by the embodiments of the present application can be realized. For details, reference may be made to the previous embodiments, which will not be elaborated here.

[0278] The above has introduced in detail a motion action evaluation method, device, system, electronic device, and storage medium provided by the embodiments of the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for evaluating a sports action, characterized in that: The steps include: Acquire motion data of the target object in the target space; Extracting motion feature data of the motion data; A motion action evaluation is performed according to the motion feature data to obtain a motion action evaluation result.

2. The method for evaluating a motor action according to claim 1, wherein: The target space includes at least two target objects; The step of extracting motion feature data of the motion data comprises: Extract motion feature data of the motion data of the at least two target objects.

3. The method for evaluating a motor action according to claim 2, wherein: The target object is an athlete and / or a sports equipment held by the athlete, and the sports characteristic data is the key element data of the athlete's body and / or the key element data of the sports equipment.

4. The method for evaluating a motor action according to claim 3, wherein: The key element data of the human body of the athlete at least includes the key element data of the human body corresponding to the athlete holding the sports equipment.

5. The method for evaluating a motor action according to claim 3, wherein: The human body key element data includes one or more combinations of the following: human body key point data, line data, and area data; the equipment key element data includes one or more combinations of the following: equipment key point data, line data, and area data.

6. The method for evaluating a motor action according to claim 5, wherein: The device key point data includes one or more combinations of the following: the highest endpoint data of the motion device, the lowest endpoint data of the motion device, the leftmost endpoint data of the motion device, the rightmost endpoint data of the motion device and / or geometric center of gravity data.

7. The method for evaluating a motor action according to any one of claims 3 to 6, characterized in that: The step of extracting motion feature data of the motion data comprises: Identifying the movement type of the target object; Extract motion feature data of the motion data according to the motion action type.

8. The method for evaluating a motor action according to claim 7, wherein: The step of performing motion evaluation according to the motion feature data to obtain a motion evaluation result comprises: Synchronously processing the key element data of the human body and the key element data of the equipment in the motion feature data; Based on the synchronously processed human body key element data and equipment key element data, motion action evaluation is performed to obtain a motion action evaluation result.

9. The method for evaluating a motor action according to claim 8, wherein: The synchronous processing of the key element data of the human body and the key element data of the equipment in the motion characteristic data includes: Performing time series modeling on the key element data of the human body and the key element data of the device to capture the time series dependency of the actions; The action boundaries are located for the key element data of the human body and the key element data of the equipment after the time series modeling, and the time series proposals for the start and end time of the action are generated; The human body key element data and the device key element data are synchronously aligned in the time dimension according to the timing proposal.

10. The method for evaluating a motor action according to claim 8, characterized in that: The step of performing motion evaluation based on the synchronously processed human key element data and the device key element data to obtain the motion evaluation result comprises: Perform time series analysis based on the synchronously processed human key element data and equipment key element data to obtain a time series analysis result; Perform correlation analysis based on the synchronously processed human key element data and equipment key element data to obtain correlation analysis results; Based on the timing analysis result and the association analysis result, a motion assessment is performed to obtain a motion assessment result.

11. The method for evaluating a motor action according to claim 8, wherein: The motion evaluation is performed based on the synchronously processed human key element data and the device key element data to obtain the motion evaluation result, including: Extracting the shared features of the human body key element data and the device key element data, and generating human body key element timing features and device key element timing features respectively; According to the time series characteristics of the key elements of the human body and the time series characteristics of the key elements of the equipment, the relative position relationship and change trend of the key element data of the human body and the key element data of the equipment in the time dimension are analyzed to obtain the movement action evaluation result.

12. The method for evaluating a motor action according to claim 10, characterized in that: The correlation analysis is performed based on the synchronously processed human key element data and the equipment key element data, and the correlation analysis results obtained include: The correlation between the human body key element data and the equipment key element data after synchronous processing at different time points is calculated to obtain the association analysis result.

13. The method for evaluating a motor action according to claim 7, characterized in that: The types of sports actions include: serve, forehand shot, two-handed backhand shot, one-handed backhand shot, forehand volley, backhand volley, forehand slice, backhand slice, and high-pressure shot.

14. The method for evaluating a motor action according to claim 3, characterized in that: After obtaining the sports action evaluation result, the method further includes: Based on the sports action evaluation results, generate personalized training suggestions.

15. The method for evaluating a motor action according to claim 14, characterized in that: The step of generating personalized training suggestions based on the sports action evaluation results includes: Acquiring the body characteristic data of the athlete and / or the equipment characteristic data of the sports equipment; Generate a real-time motion trajectory of the athlete based on the body shape feature data and / or the equipment characteristic data; Comparing the real-time motion trajectory with the standard motion trajectory of the athlete at the corresponding level to obtain a motion trajectory comparison result; Generate personalized training suggestions for the athlete based on the sports action evaluation results and the sports trajectory comparison results.

16. The method for evaluating a motor action according to claim 15, characterized in that: The method further comprises: Obtaining historical sports action evaluation results of athletes, and generating corresponding training technology reports based on the historical sports action evaluation results; At least one of the sports action evaluation result, the personalized training suggestion and the training technology report is fed back to the athlete's smart mobile terminal in real time.

17. A sports action evaluation device, characterized in that: include: A data acquisition module, used to acquire motion data of a target object in a target space; A feature extraction module, used to extract motion feature data of the motion data; The motion evaluation module is used to perform motion evaluation according to the motion feature data to obtain a motion evaluation result.

18. A sports action evaluation system, characterized in that: Including data acquisition equipment, tennis service equipment and display equipment; The data acquisition device is used to obtain the motion data of the target object in the target space and send it to the tennis service device; The tennis service device is used to extract motion feature data of the motion data, and perform motion action evaluation according to the motion feature data to obtain a motion action evaluation result; The display device is used to provide real-time feedback of the sports action evaluation result.

19. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the motor action evaluation method as described in any one of claims 1 to 16 when executing the computer program.

20. A storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the motor action evaluation method as described in any one of claims 1 to 16.

Images