Fitness assisting method and device, equipment, storage medium and product

By decoding and key point detection of fitness videos, generating dynamic models, identifying and correcting wrong movements, the problem that existing fitness methods and equipment cannot improve fitness results is solved, safe and scientific fitness guidance is achieved, and training efficiency is improved. Effect.

CN119028012APending Publication Date: 2024-11-26SHENZHEN JIUZHOU ELECTRIC
View PDF 0 Cites 8 Cited by

Patent Information

Application Number
CN202411007580.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing fitness methods and equipment cannot effectively improve the fitness effect, and the level of personal training guidance is uneven, posing safety hazards, and the price is high, which many people cannot afford.

Method used

By decoding and frame extraction of user fitness videos, detecting key points of human body, generating dynamic models, and pose recognition based on preset pose recognition models. If wrong actions are found, corrective data will be generated to provide real-time feedback and guidance.

Benefits of technology

It realizes refined analysis and real-time monitoring of user movements, discovers subtle errors and potential risks, improves training efficiency, ensures that users conduct effective training under safe and scientific guidance, and avoids sports injuries.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119028012A_ABST
    Figure CN119028012A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of artificial intelligence, in particular to an auxiliary fitness method and device, equipment, a storage medium and a product. The method comprises the following steps: performing decoding and frame extraction on a user fitness video to obtain fitness continuous frames; performing human body key point detection on each frame in the fitness continuous frames to obtain key point information; generating a human body dynamic model based on the key point information; based on a preset posture recognition model, performing posture recognition on the human body dynamic model to obtain a motion posture set; and if the wrong motion posture exists in the motion posture set, generating correction data based on a standard posture corresponding to the wrong motion posture in a preset standard posture library. According to the method, refined decoding and frame extraction are carried out on the fitness video of the user, the key points of the human body can be accurately detected, the dynamic model is generated, then the preset posture recognition model is used for recognizing the motion posture and correcting wrong actions in time, and therefore the fitness effect is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an auxiliary fitness method, device, equipment, storage medium and product. Background Art

[0002] With the improvement of living standards, people are paying more and more attention to their health, and the number of people doing fitness is also increasing. However, many people's fitness postures and methods are not scientific, the fitness results are not ideal, and there are even certain hidden dangers. In order to exercise correctly and scientifically, more and more people choose to hire private trainers, but the guidance level of private trainers varies, the results are not necessarily good, and the price of private trainers is high, and many people cannot afford it. Therefore, how to improve fitness results has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The main purpose of this application is to provide an auxiliary fitness method, device, equipment, storage medium and product, aiming to solve the technical problem of how to improve fitness effects.

[0004] To achieve the above objectives, the present application provides a method for assisting fitness, the method comprising:

[0005] Decode and extract frames of user fitness videos to obtain continuous fitness frames;

[0006] Performing human body key point detection on each of the continuous fitness frames to obtain key point information;

[0007] Based on the key point information, a human body dynamic model is generated;

[0008] Based on a preset posture recognition model, performing posture recognition on the human body dynamic model to obtain a motion posture set;

[0009] If there is an erroneous motion gesture in the motion gesture set, correction data is generated based on a standard gesture corresponding to the erroneous motion gesture in a preset standard gesture library.

[0010] In one embodiment, the step of generating a human body dynamic model based on the key point information includes:

[0011] Based on a preset depth estimation algorithm, converting the key point information into three-dimensional coordinate information;

[0012] Based on the three-dimensional coordinate information, a human skeleton model is generated;

[0013] The human body dynamic model is generated based on the key point information and the human body skeleton model.

[0014] In one embodiment, the step of performing posture recognition on the human body dynamic model based on a preset posture recognition model to obtain a motion posture set includes:

[0015] Determine the posture frame sequence according to the human body dynamic model;

[0016] Extracting features from the posture frame sequence, and inputting the feature extraction results into the preset posture recognition model;

[0017] The motion gesture set is generated based on the output result of the preset gesture recognition model.

[0018] In one embodiment, after the step of detecting human key points on each of the continuous fitness frames to obtain key point information, the method further includes:

[0019] Determine the changes in the bone structure according to the angle changes and speeds of each joint in the key point information;

[0020] Based on the changes in the bone structure and a preset dangerous action detection model, a dangerous level assessment is performed on the current action;

[0021] Based on the evaluation results, corresponding warning information is generated and pushed to the user's device.

[0022] In one embodiment, after the step of performing posture recognition on the human body dynamic model based on a preset posture recognition model to obtain a motion posture set, the method further includes:

[0023] Based on a preset action classification model, classifying the motion postures in the motion posture set into multiple types;

[0024] Based on the preset energy consumption model, determine the energy consumption per unit time of each type of motion posture;

[0025] Determine the total energy consumption according to the type of each motion posture in the motion posture set and the corresponding motion duration;

[0026] Adjust fitness plans and dietary recommendations based on the user's training goal data and the total energy consumption.

[0027] In one embodiment, after the step of determining the energy consumption per unit time of each type of motion posture based on the preset energy consumption model, the method further includes:

[0028] The total amount of energy consumption and the duration of the motion posture set are used as actual motion data;

[0029] Comparing the actual exercise data with the preset target data, and judging whether the user is slacking off based on the comparison result;

[0030] If yes, send a slack reminder message to the user's device.

[0031] In addition, to achieve the above-mentioned purpose, the present application also proposes an auxiliary fitness device, which includes:

[0032] A preprocessing module is used to decode and extract frames from user fitness videos to obtain fitness continuous frames;

[0033] A key point detection module, used to detect the key points of the human body in each of the continuous fitness frames to obtain key point information;

[0034] A dynamic model generation module, used to generate a human body dynamic model based on the key point information;

[0035] A posture recognition module, used to perform posture recognition on the human body dynamic model based on a preset posture recognition model to obtain a motion posture set;

[0036] The target correction module is used to generate correction data based on a standard posture corresponding to the erroneous motion posture in a preset standard posture library if there is an erroneous motion posture in the motion posture set.

[0037] In addition, to achieve the above objectives, the present application also proposes an auxiliary fitness device, which includes: a memory, a processor, and an auxiliary fitness program stored in the memory and executable on the processor, wherein the auxiliary fitness program is configured to implement the steps of the auxiliary fitness method described above.

[0038] In addition, to achieve the above objectives, the present application also proposes a storage medium, on which an auxiliary fitness program is stored, and when the auxiliary fitness program is executed by a processor, the steps of the auxiliary fitness method described above are implemented.

[0039] In addition, to achieve the above-mentioned purpose, the present application also proposes a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the auxiliary fitness method described above are implemented.

[0040] The present application obtains fitness continuous frames by decoding and extracting frames of user fitness videos; performs human key point detection on each frame in the fitness continuous frames to obtain key point information; generates a human dynamic model based on the key point information; performs posture recognition on the human dynamic model based on a preset posture recognition model to obtain a motion posture set; if there is an erroneous motion posture in the motion posture set, generates correction data based on the standard posture corresponding to the erroneous motion posture in the preset standard posture library. The present application can finely analyze the movement of each moment to ensure the accuracy of the analysis by decoding and extracting frames of user fitness videos; accurately locate the joint position by using human key point detection technology to provide data support for generating dynamic models; the dynamic model fully reflects the user's motion posture, so that the system can monitor the movement in all directions and find subtle errors and potential risks. The preset posture recognition model can automatically analyze the posture, feedback errors in real time, and improve training efficiency. Finally, the correction data generated by the standard posture library can scientifically and accurately guide users to correct wrong movements and avoid sports injuries, thereby greatly improving the fitness effect and ensuring that users can conduct effective training under safe and scientific guidance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flow chart of the first embodiment of the auxiliary fitness method of the present application;

[0042] Figure 2 This is a schematic diagram of a sub-process in the second embodiment of the auxiliary fitness method of this application;

[0043] Figure 3 This is a schematic diagram of a sub-process in the third embodiment of the auxiliary fitness method of this application;

[0044] Figure 4 This is a schematic diagram of the module structure of the auxiliary fitness device according to an embodiment of the present application;

[0045] Figure 5 This is a schematic diagram of the device structure of the hardware operating environment involved in the auxiliary fitness method in the embodiment of the present application.

[0046] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0047] 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.

[0048] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0049] It should be noted that with the improvement of living standards, people are paying more and more attention to their physical health, and the number of people doing fitness is also increasing. However, many people's fitness postures and methods are not scientific, the fitness results are not ideal, and there are even certain hidden dangers. In order to exercise correctly and scientifically, more and more people choose to hire private trainers, but the guidance level of private trainers varies, the results are not necessarily good, and the price of private trainers is high, and many people cannot afford it. Therefore, how to improve fitness results has become a technical problem that needs to be solved urgently.

[0050] The main solution of the present application is: by decoding and extracting frames of user fitness videos, fitness continuous frames are obtained; human key point detection is performed on each frame in the fitness continuous frames to obtain key point information; based on the key point information, a human dynamic model is generated; based on a preset posture recognition model, posture recognition is performed on the human dynamic model to obtain a motion posture set; if there is an erroneous motion posture in the motion posture set, correction data is generated based on a standard posture corresponding to the erroneous motion posture in a preset standard posture library.

[0051] This application can analyze the movement of each moment in detail and ensure the accuracy of the analysis by decoding and extracting frames from the user's fitness video; it uses human key point detection technology to accurately locate the joint position and provide data support for generating a dynamic model; this dynamic model fully reflects the user's movement posture, enabling the system to monitor the movement in all directions and detect subtle errors and potential risks. The preset posture recognition model can automatically analyze postures, provide real-time feedback on errors, and improve training efficiency. Ultimately, the correction data generated by the standard posture library can scientifically and accurately guide users to correct incorrect movements and avoid sports injuries, thereby greatly improving fitness results and ensuring that users can conduct effective training under safe and scientific guidance.

[0052] It should be noted that the execution subject of the method of this embodiment can be a computing service device with data processing, network communication and program running functions, or it can be the above-mentioned auxiliary fitness equipment with the same or similar functions. This embodiment and the following embodiments will be described by taking the auxiliary fitness equipment as an example.

[0053] Based on this, the first embodiment of the auxiliary fitness method of this application is proposed, please refer to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the auxiliary fitness method of the present application.

[0054] In this embodiment, the method includes the following steps:

[0055] S1: Decode and extract frames of the user's fitness video to obtain continuous fitness frames;

[0056] It should be noted that decoding refers to converting compressed and coded video data into uncompressed frame data for further processing. Frame extraction refers to extracting continuous image frames from a video stream, each frame representing a moment in the video.

[0057] Specifically, receive fitness videos uploaded by users or videos shot in real time by cameras. Call the video decoding library (such as FFmpeg, OpenCV, etc.), initialize the decoder, and prepare to process video data. Pass the video data stream to the decoder, which parses the video file format and decodes the compressed video data into uncompressed image frames. The decoded image frames are temporarily stored in the memory for subsequent processing.

[0058] Furthermore, the frequency of frame extraction is set as needed (for example, 30 frames per second) to ensure that sufficiently detailed continuous actions are obtained. The images are read frame by frame from the decoded video stream and each frame is extracted. A number and a timestamp are assigned to each frame image to ensure that the action can be accurately analyzed in chronological order during subsequent processing. The extracted frame images are stored in a specified storage location (such as a memory cache or hard disk) to provide raw data for subsequent key point detection and dynamic model generation.

[0059] Through high-frequency frame extraction, every tiny movement of the user can be captured to ensure that no details are missed. This is especially important for subsequent movement analysis and correction. Decoding and frame extraction provide high-quality raw data, so that subsequent key point detection and posture analysis can be based on accurate image information. This helps to improve the accuracy and reliability of the overall analysis. Decoding and frame extraction of real-time video can achieve real-time monitoring and analysis, and timely feedback on user movement problems. This improves the training effect and safety of users. Through the video decoding step, it can be compatible with fitness videos of different formats, providing flexibility and wide applicability. Frame extraction provides basic data for subsequent steps (such as key point detection and 3D modeling), ensuring the smooth and efficient operation of subsequent processing steps.

[0060] S2: Detect human key points on each of the continuous fitness frames to obtain key point information;

[0061] It should be noted that fitness continuous frames refer to continuous image frames extracted from the video, and each frame represents a moment in the video. Key point detection refers to identifying and locating the main joints and parts of the human body in the image (such as head, shoulders, elbows, wrists, hips, knees, ankles, etc.).

[0062] Specifically, select a pre-trained key point detection model, such as OpenPose, HRNet, or MediaPipe. These models have been trained on large-scale human key point datasets and have high detection accuracy. Optionally, load a pre-trained model in the program, including the model's weights and structure, for key point detection. Scale each frame of the image to the input size required by the model and normalize it (such as normalizing the pixel values ​​to the [0,1] interval) to match the model's input standard. Convert the image to the data format required by the model (such as a tensor) and prepare it for input into the model for detection.

[0063] Furthermore, the preprocessed image is input into the key point detection model. The model processes the input image and outputs the coordinates of each detected key point (usually two-dimensional coordinates x, y). Each key point coordinate is also accompanied by a confidence score, which indicates the model's confidence in the prediction of the key point location. Based on the confidence score, low-confidence key points are filtered out to improve the accuracy of the detection. Based on the detected key point coordinates, the corresponding joints are connected to form a human skeleton diagram for subsequent posture analysis. The key point information of each frame (including coordinates and confidence scores) is stored in a data structure, which usually includes a dictionary or array to save the name, coordinates and confidence of each key point. The key point information is associated with the corresponding timestamp or frame number for subsequent timing analysis.

[0064] Through key point detection, the position of each joint and limb can be accurately located, providing an accurate data basis for subsequent posture analysis and 3D modeling. The detected key point information can be used to analyze the user's motion posture, identify incorrect postures and provide correction suggestions to improve the user's fitness effect. Key point detection can quickly process frame images, realize real-time monitoring and feedback, correct the user's movements in time, and avoid sports injuries. Key point information can be used to generate a human skeleton diagram to intuitively display the user's motion state, which helps users understand and improve their movements. Key point detection is not only suitable for fitness scenarios, but also for other applications that require human posture analysis, such as sports training, rehabilitation therapy, virtual reality, etc. Human key point detection is performed on each frame in the fitness continuous frame, and the user's joints and limbs are accurately located, providing an accurate data basis for posture analysis and 3D modeling. This step can not only accurately analyze and feedback the user's motion posture, help users correct incorrect movements in time and avoid sports injuries, but also generate intuitive visual data to improve the user's fitness effect. Through real-time monitoring and analysis, efficient fitness guidance and improvement suggestions can be provided.

[0065] S3: Generate a human body dynamic model based on the key point information;

[0066] It should be noted that the key point information refers to the coordinates and confidence scores of the detected human joints. The human body dynamic model represents a three-dimensional model of the human body in motion, which can dynamically display the human body posture and movement.

[0067] Specifically, the key point coordinates and confidence scores of each frame are obtained from the key point detection step. The key point information in consecutive frames is sorted in chronological order to form time series data. The two-dimensional key point coordinates are converted into three-dimensional coordinates using a preset depth estimation algorithm (such as a deep learning model based on monocular vision). The algorithm can infer the depth information of each key point by learning a large amount of motion data. The three-dimensional coordinates of each key point are generated by combining the two-dimensional coordinates and the estimated depth information.

[0068] Furthermore, according to the human anatomical structure, key points (such as head, shoulder, elbow, wrist, hip, knee, ankle, etc.) are connected into a skeleton. A connection matrix or a connection list is usually used to define the connection relationship of the joint points. The three-dimensional coordinates are used through graphics rendering technology to generate a visual three-dimensional skeleton model. The connection lines between each joint point can intuitively show the posture and movement of the human body. The skeleton models of consecutive frames are combined in time sequence to form a dynamic model. The dynamic model can show the posture changes of the human body in a time series.

[0069] Optionally, smooth the key point data between consecutive frames to eliminate noise and jitter to ensure smooth transition and realism of the dynamic model. Store the generated dynamic model data in a server or local storage in a standard 3D model file (such as OBJ, FBX, etc.) or a custom format. Use a graphics engine (such as OpenGL, Unity, UnrealEngine, etc.) to render the dynamic model in real time and display it on the user's display device (such as a mobile phone, PC, smart mirror, etc.).

[0070] The human body dynamic model displays human body posture through three-dimensional coordinates, allowing users to observe their movements in three-dimensional and intuitive ways, understand and improve training results. Based on the three-dimensional dynamic model, more accurate posture analysis can be performed to detect subtle errors in complex movements. The human body dynamic model can be updated and rendered in real time, providing instant feedback to help users quickly correct incorrect movements and improve training efficiency. The three-dimensional dynamic model provides rich visual data to help users understand their own movement postures and compare them with standard postures. Through the three-dimensional dynamic model, potential sports injuries can be more effectively detected and prevented, and the safety of training can be improved. The dynamic model provides a data basis for personalized fitness guidance, and can provide tailored training suggestions based on the user's specific posture and movements. The human body dynamic model is generated based on key point information, and a three-dimensional model that can dynamically display human body posture and movement is generated through three-dimensional coordinate transformation and skeleton model construction. This step not only provides three-dimensional and intuitive movement display, but also can perform accurate posture analysis and real-time feedback, helping users understand and improve training effects, improve training efficiency and safety, and provide data support for personalized fitness guidance.

[0071] S4: Based on a preset posture recognition model, performing posture recognition on the human body dynamic model to obtain a motion posture set;

[0072] It should be noted that the preset posture recognition model is a pre-set pre-trained AI model used to recognize and classify different motion postures of the human body. The human body dynamic model represents a three-dimensional model of the human body in motion, which can dynamically display the human body posture and movement. The motion posture set is a collection of various recognized motion postures.

[0073] Specifically, select a suitable preset posture recognition model, such as a model based on a convolutional neural network (CNN) or a long short-term memory network (LSTM). These models have been trained on a large amount of human posture data. Load the weights and structure of the preset model in the program and prepare for posture recognition. Extract continuous frames from the dynamic model to form a posture frame sequence. Each frame contains the three-dimensional coordinate information of the joint point. Normalize the frame sequence to ensure that the input data meets the model requirements (such as coordinate normalization, time series normalization).

[0074] Further, features are extracted from the posture frame sequence, such as the distance between joints, angle changes, speed and acceleration. Feature engineering helps improve the recognition accuracy of the model. The extracted features are converted into the input format required by the model, such as a two-dimensional array or tensor. The preprocessed feature data is input into the posture recognition model. The model classifies the posture of each frame or each sequence according to the input feature data, and outputs the posture category (such as squat, push-up, lunge, etc.). The model also outputs the confidence score of each posture classification, indicating the confidence of the model in the classification result. The posture categories and confidence scores output by the model are integrated in chronological order to form a motion posture set. The motion posture set contains the type, start time and end time of each recognized posture. The generated motion posture set is stored in a database or file system for subsequent analysis and processing.

[0075] Through the preset posture recognition model, the system can accurately identify the user's various motion postures, providing a reliable basis for subsequent motion analysis and correction. The posture recognition process can be carried out in real time, providing instant posture analysis and feedback, helping users to quickly improve their movements and improve training efficiency. The preset model has been trained with a large amount of data and can recognize a variety of common and complex motion postures, with wide coverage and flexible application. The posture recognition model automatically processes data, reduces manual intervention, and improves analysis efficiency and accuracy. Based on the identified motion posture set, the system can provide personalized training suggestions and guidance to improve user experience and training effectiveness. Based on the preset posture recognition model, posture recognition is performed on the human body dynamic model to generate a motion posture set. By loading the preset model, extracting the posture frame sequence features, and inputting them into the model for classification, various motion postures can be accurately identified. This step not only provides real-time posture analysis and feedback, but also comprehensively covers a variety of motion types, realizes automated analysis and personalized guidance, and improves the user's training efficiency and effectiveness.

[0076] S5: If there is an erroneous motion posture in the motion posture set, generating correction data based on a standard posture corresponding to the erroneous motion posture in a preset standard posture library.

[0077] It should be noted that the preset standard posture library refers to a pre-set database containing various correct motion postures for comparison and correction of incorrect postures. Correction data refers to information used to guide users to correct incorrect motion postures, which may include correction tips, correct posture examples, etc.

[0078] Specifically, each posture in the motion posture set is analyzed to detect whether it meets the standard of the correct posture. For example, whether the joint position, angle or motion trajectory of a certain posture is within the allowable error range. If a posture deviates from the standard posture, it is marked as an incorrect posture, and the specific error information of the posture is recorded (such as excessive joint angle, insufficient movement range, etc.). The incorrect posture is compared with the posture in the standard posture library to find the corresponding standard posture. Features such as joint position, angle and motion trajectory can be used for comparison. Calculate the matching degree between the incorrect posture and the standard posture, and select the standard posture with the highest matching degree as a reference for correction.

[0079] Furthermore, the differences between the incorrect posture and the corresponding standard posture are analyzed, including differences in joint position, angle, motion trajectory, etc. Based on the difference analysis results, specific correction prompts are generated. For example, the user is prompted to what position a joint needs to be adjusted to, the range of motion needs to be increased or decreased, etc. An example video or animation of the correct posture can be generated to intuitively display the standard posture to help users understand and imitate. The correction prompts and examples are integrated into a complete correction data package. The correction data is pushed to the user through mobile devices (such as mobile phones, smart watches, fitness mirrors, etc.) to provide real-time guidance. After the user adjusts the movement according to the correction prompts, the posture is detected again to confirm whether the correction is successful, and further guidance is provided if necessary.

[0080] By comparing the standard posture library with the wrong posture, the system can generate correction data in a targeted manner to help users accurately correct their movements. By generating and pushing correction data in real time, users can get immediate feedback and guidance, quickly adjust their postures, and avoid the accumulation of wrong movements. Correction example videos or animations provide intuitive visual references, so that users can more easily understand and imitate correct movements and improve the correction effect. Through multiple tests and feedback, users can gradually improve their movements, form correct exercise habits, and improve the quality and effect of training. Correct wrong movements in a timely manner to reduce sports injuries caused by improper movements and ensure the safety of users. Based on the preset standard posture library, the wrong postures that exist in the sports posture collection are compared and analyzed to generate specific correction data. By identifying wrong postures, finding corresponding standard postures, analyzing differences, generating and pushing correction prompts and examples, users can be accurately guided to correct wrong movements. This process not only provides real-time and intuitive correction guidance to help users quickly adjust and improve their postures, but also improves the quality and safety of training and promotes the formation of correct exercise habits.

[0081] This embodiment obtains fitness continuous frames by decoding and extracting frames of user fitness videos; performs human key point detection on each frame in the fitness continuous frames to obtain key point information; generates a human dynamic model based on the key point information; performs posture recognition on the human dynamic model based on a preset posture recognition model to obtain a motion posture set; if there is an erroneous motion posture in the motion posture set, generates correction data based on the standard posture corresponding to the erroneous motion posture in the preset standard posture library. This embodiment can finely analyze the movement of each moment to ensure the accuracy of the analysis by decoding and extracting frames of user fitness videos; uses human key point detection technology to accurately locate the joint position and provide data support for generating a dynamic model; the dynamic model fully reflects the user's motion posture, so that the system can monitor the movement in all directions and find subtle errors and potential risks. The preset posture recognition model can automatically analyze the posture, feedback errors in real time, and improve training efficiency. Finally, the correction data generated by the standard posture library can scientifically and accurately guide users to correct erroneous movements and avoid sports injuries, thereby greatly improving the fitness effect and ensuring that users can conduct effective training under safe and scientific guidance.

[0082] Based on the above first embodiment, a second embodiment of the fitness assistance method of the present application is proposed. Figure 2 , Figure 2 This is a schematic diagram of a sub-process in the second embodiment of the auxiliary fitness method of the present application.

[0083] like Figure 2 As shown, in this embodiment, step S3 includes:

[0084] S31: based on a preset depth estimation algorithm, converting the key point information into three-dimensional coordinate information;

[0085] S32: generating a human skeleton model based on the three-dimensional coordinate information;

[0086] S33: Generate the human body dynamic model based on the key point information and the human body skeleton model.

[0087] It should be noted that the preset depth estimation algorithm refers to a preset algorithm for converting two-dimensional key point information into three-dimensional coordinate information. The three-dimensional coordinate information refers to the three-dimensional position coordinates (x, y, z) of each key point. The skeleton model is a three-dimensional structure formed by the connection between key points, which represents the posture and skeleton of the human body.

[0088] Specifically, select and load a preset depth estimation algorithm, such as a deep learning model based on monocular vision or a multi-view reconstruction algorithm. Input the coordinates of the two-dimensional key points in each frame into the depth estimation algorithm. The algorithm infers the depth information of each key point based on the two-dimensional position of the key point and the known motion characteristics, and generates three-dimensional coordinates (x, y, z). Output the three-dimensional coordinate information of each key point as the basis for subsequent skeleton model construction.

[0089] Furthermore, according to the human anatomical structure, the key points in the three-dimensional coordinate information are connected according to the preset connection rules to form a three-dimensional skeleton. The connection rules usually define which joints are directly connected (such as shoulders and elbows, knees and ankles, etc.). Using graphics rendering technology (such as OpenGL or DirectX), the connected skeleton model is rendered to generate a three-dimensional skeleton graph. The skeleton model of each frame is arranged in chronological order to form time series data. The three-dimensional coordinates between consecutive frames are smoothed to eliminate noise and jitter to ensure a smooth transition of the dynamic model. The smoothed time series data is rendered in real time through a graphics engine (such as Unity, UnrealEngine, etc.) to generate a dynamic three-dimensional human body model to show the movement process of the human body.

[0090] Through the three-dimensional coordinates and skeleton model, the user's motion posture can be displayed in three dimensions, allowing users to observe and understand their movements more intuitively. The three-dimensional coordinates provide richer position information than the two-dimensional coordinates, making posture analysis and error detection more accurate. The dynamic model can update and display the user's motion process in real time, provide instant feedback, and help users quickly improve their movements. Through the detailed three-dimensional skeleton and dynamic model, users can better adjust and optimize their motion posture and improve the training effect. Timely discover and correct wrong movements, reduce the risk of injury that may occur during exercise, and ensure user safety. Based on the preset depth estimation algorithm, the key point information is converted into three-dimensional coordinate information, and these three-dimensional coordinates are used to generate a three-dimensional skeleton model and a dynamic model. This process enables the user's motion posture and dynamic process to be displayed in three dimensions by loading the depth estimation algorithm, generating three-dimensional coordinates, building and rendering the skeleton model and the dynamic model. Through high-precision three-dimensional analysis and real-time feedback, help users optimize their movements and improve training effects and safety.

[0091] Based on the above first embodiment, in this embodiment, step S4 includes:

[0092] S41: determining a posture frame sequence according to a human body dynamic model;

[0093] S42: extracting features from the posture frame sequence, and inputting the feature extraction results into the preset posture recognition model;

[0094] S43: Generate the motion gesture set based on the output result of the preset gesture recognition model.

[0095] It should be noted that the posture frame sequence refers to a human posture data sequence arranged in chronological order. Feature extraction refers to extracting key features for posture recognition from the posture frame sequence.

[0096] Specifically, continuous frames are extracted from the human body dynamic model to form time series data. Each frame is assigned a number and a timestamp to ensure the order and integrity of the data. The continuous frames are arranged in chronological order to form a posture frame sequence for subsequent feature extraction and posture recognition. The three-dimensional coordinate information (x, y, z) of the key points is extracted from each frame. The relative angles between the joints are calculated, especially the bending angles of the joints such as the elbow and knee. The speed and acceleration of the joints are calculated to reflect the speed and changes of the movements. The above features are arranged in chronological order to form a feature sequence to ensure the data consistency of each frame.

[0097] Furthermore, the extracted feature sequence is standardized to meet the input requirements of the gesture recognition model. The preprocessed feature sequence is input into the preset gesture recognition model. The model classifies the gestures according to the input feature sequence and identifies the type of motion gesture corresponding to each frame. The output results of the gesture recognition model are integrated in chronological order to form a complete motion gesture set. Each recognized gesture is labeled and its type, start time and end time are recorded. The generated motion gesture set is stored in a database or file system for subsequent analysis and processing.

[0098] Through efficient feature extraction and posture recognition models, the system can accurately identify the user's various motion postures, providing a reliable basis for subsequent motion analysis and correction. The posture recognition process can be carried out in real time, providing instant posture analysis and feedback, helping users to quickly improve their movements and improve training efficiency. The preset model has been trained with a large amount of data and can recognize a variety of common and complex motion postures, with wide coverage and flexible application. The posture recognition model automatically processes data, reduces manual intervention, and improves analysis efficiency and accuracy. Based on the identified motion posture set, the system can provide personalized training suggestions and guidance to improve user experience and training results. Based on the human body dynamic model, the posture frame sequence is first generated and feature extraction is performed, and then the features are input into the preset posture recognition model for posture classification, and finally a motion posture set is generated. Through efficient feature extraction and posture recognition, the system can accurately and real-time identify the user's various motion postures, providing a reliable data basis for personalized guidance and training optimization, thereby improving the user's training effect and safety.

[0099] This embodiment obtains fitness continuous frames by decoding and extracting frames of user fitness videos; performs human key point detection on each frame in the fitness continuous frames to obtain key point information; generates a human dynamic model based on the key point information; performs posture recognition on the human dynamic model based on a preset posture recognition model to obtain a motion posture set; if there is an erroneous motion posture in the motion posture set, generates correction data based on the standard posture corresponding to the erroneous motion posture in the preset standard posture library. This embodiment can finely analyze the movement of each moment to ensure the accuracy of the analysis by decoding and extracting frames of user fitness videos; uses human key point detection technology to accurately locate the joint position and provide data support for generating a dynamic model; the dynamic model fully reflects the user's motion posture, so that the system can monitor the movement in all directions and find subtle errors and potential risks. The preset posture recognition model can automatically analyze the posture, feedback errors in real time, and improve training efficiency. Finally, the correction data generated by the standard posture library can scientifically and accurately guide users to correct erroneous movements and avoid sports injuries, thereby greatly improving the fitness effect and ensuring that users can conduct effective training under safe and scientific guidance.

[0100] Based on the above second embodiment, a third embodiment of the fitness assistance method of the present application is proposed. Figure 3 , Figure 3 This is a schematic diagram of a sub-process in the third embodiment of the auxiliary fitness method of the present application.

[0101] In this embodiment, after step S2, the method further includes:

[0102] S2a: determining the changes in the bone structure according to the angle changes and speeds of each joint in the key point information;

[0103] S2b: Based on the changes in the bone structure and a preset dangerous action detection model, assessing the danger level of the current action;

[0104] S2c: Based on the evaluation results, a corresponding warning message is generated and pushed to the user's device.

[0105] It should be noted that changes in bone structure refer to changes in motion state reflected by joint angles, speeds, etc. Dangerous action detection models are pre-trained models used to assess the danger level of actions. Warning information refers to notifications that prompt users that their actions are potentially dangerous.

[0106] Specifically, the joint angles in each frame are calculated based on the key point information. For example, the bending angles of joints such as elbows and knees are calculated using trigonometric functions. The speed of each joint point is calculated, including linear speed and angular speed. The speed change is calculated through time intervals. The joint angles and speeds of each frame are recorded to form time series data to reflect the changes in bone structure during movement. Select and load a preset dangerous action detection model. The model can be based on machine learning algorithms such as decision trees, random forests, or neural networks, and is trained with a large amount of data. The features of the changes in bone structure (such as joint angles, speeds, accelerations, etc.) are input into the dangerous action detection model. The focus of feature extraction is on key indicators that can reflect the danger of the action. Based on the input feature data, the model evaluates the danger level of the current action and outputs a danger score or classification result (such as low risk, medium risk, high risk).

[0107] Furthermore, the danger level of the current action is analyzed based on the output results of the dangerous action detection model. If the evaluation results show that the action is highly dangerous, the system needs to generate a corresponding warning message. Specific warning information is generated based on the danger level, including the cause of the danger, the details of the action that need to be adjusted, possible consequences, etc. For example, the user is prompted that "the current action has a too large knee bending angle and there is a risk of injury, please adjust your posture." The generated warning information is pushed to the user through a mobile device (such as a mobile phone, smart watch, fitness mirror, etc.) to provide real-time feedback and guidance.

[0108] The system can detect and evaluate the user's movement status in real time, detect potential dangerous movements in time, and provide warnings to avoid sports injuries. Through accurate analysis of changes in bone structure, the system can provide detailed movement adjustment suggestions to help users improve the quality of their movements. Real-time monitoring and warning functions significantly improve the safety of users during exercise and reduce the risk of injury caused by improper movements. Timely and detailed warning information and guidance enhance the user's training experience, allowing users to train more efficiently with peace of mind. Based on data-driven analysis of key point information and bone structure changes, the system can continuously optimize the dangerous movement detection model to improve the accuracy and reliability of evaluation. According to the angle changes and speeds of each joint in the key point information, the system can determine the changes in bone structure and use the preset dangerous movement detection model to evaluate the danger level. Based on the evaluation results, the system generates corresponding warning information and pushes it to the user's device. This process uses real-time and accurate dangerous movement detection and feedback to timely warn of potential dangers, provide detailed guidance, improve the user's training safety and effectiveness, and enhance the overall training experience.

[0109] Based on the above second embodiment, in this embodiment, after step S4, the following is further included:

[0110] S4a: Classifying the motion postures in the motion posture set into multiple types based on a preset motion classification model;

[0111] S4b: Determine the energy consumption per unit time of each type of motion posture based on a preset energy consumption model;

[0112] S4c: determining the total energy consumption according to the type of each motion posture in the motion posture set and the corresponding motion duration;

[0113] S4d: Adjusting fitness plans and dietary recommendations based on the user's training target data and the total energy consumption.

[0114] It should be noted that the preset action classification model is a pre-set pre-trained AI model used to classify motion postures into different types of motion. The preset energy consumption model is a model used to calculate the energy consumption per unit time of different motion postures. The total energy consumption refers to the total energy consumption of all motion postures over a period of time. The training target data includes the fitness goals set by the user, such as weight loss, muscle gain, etc.

[0115] Specifically, a preset motion classification model is selected and loaded, such as a model based on a convolutional neural network (CNN) or a long short-term memory network (LSTM). Each posture in the motion posture set is input into the motion classification model. The model classifies each posture into a predefined motion type (such as running, squats, push-ups, etc.) based on the input posture data. The classification results are stored in chronological order to form a classified motion posture set, and the type, start time and end time of each posture are recorded. Select and load a preset energy consumption model, which is usually based on MET (metabolic equivalent) values. According to each type of exercise and the corresponding MET value, the energy consumption per unit time of each motion posture is calculated. For example, the MET value of running is 8, and the calculation formula is: consumption = MET value * body weight (kg) * time (hours). The calculated energy consumption is stored in chronological order to form an energy consumption data set.

[0116] Furthermore, the energy consumption of all movement postures is integrated according to the classification results of the movement postures and the duration of each posture. The energy consumption of all movement postures is accumulated to obtain the total energy consumption over a period of time. The user's training target data (such as weight loss goals, muscle gain goals, etc.) is compared with the actual total energy consumption. According to the comparison results, the user's fitness plan is adjusted. For example, if the actual energy consumption is lower than the target, increase the training intensity or time; if the energy consumption is too high, reduce the training volume appropriately. According to the total energy consumption and the user's training goals, provide corresponding dietary recommendations, including the daily required calories, protein, carbohydrate and fat intake, as well as specific recipe recommendations. According to user feedback and progress, dynamically adjust the fitness plan and dietary recommendations to ensure that users can continue to achieve the set fitness goals.

[0117] Through the action classification model, the system can accurately classify different types of sports postures, providing an accurate data basis for subsequent energy consumption calculations. Using the capacity consumption model, the system can scientifically calculate the energy consumption of each sports posture to help users understand the energy consumption of each training. According to the user's training goals and actual energy consumption data, the system can dynamically adjust the fitness plan and diet recommendations to ensure that the user effectively achieves the fitness goal. The personalized fitness plan and diet recommendations provided by the system can be adjusted according to the user's actual situation and goals to improve the user experience and training effect. Through continuous monitoring and analysis of user training data, the system can continuously optimize the fitness plan and diet recommendations to improve the overall training effect. Based on the preset action classification model, the system classifies different types of sports postures in the sports posture set, and uses the capacity consumption model to calculate the energy consumption per unit time of each type of sports posture. According to the classified sports postures and the corresponding duration, the total energy consumption of the user is determined. Based on the user's training target data, the system dynamically adjusts the fitness plan and diet recommendations to ensure that the user can effectively achieve the set fitness goals. This process improves the user's training effect and safety and optimizes the overall fitness experience through accurate classification, scientific energy consumption calculation and personalized adjustment.

[0118] Based on the above second embodiment, in this embodiment, after step S4b, the following is further included:

[0119] S4ba: taking the total energy consumption and the duration of the motion posture set as actual motion data;

[0120] S4bb: comparing the actual exercise data with the preset target data, and judging whether the user is slacking off based on the comparison result;

[0121] S4bc: If yes, send a slack reminder message to the user device.

[0122] It should be noted that actual exercise data refers to data including total energy consumption and duration of exercise posture sets. Preset target data refers to fitness target data set by the user, such as target energy consumption, target exercise time, etc. Slack reminder is a reminder message sent when it is detected that the user has not reached the preset target.

[0123] Specifically, the energy consumption of each exercise posture is accumulated to obtain the total energy consumption over a period of time. The duration of each exercise posture is recorded, and the total exercise duration is calculated. The total energy consumption and exercise duration are integrated into actual exercise data for comparison with preset target data. Preset target data is obtained from the fitness plan set by the user, such as daily or weekly target energy consumption and target exercise time. The actual exercise data is compared with the preset target data to calculate the difference between the actual completion status and the target. For example, whether the actual energy consumption reaches the target energy consumption, and whether the actual exercise duration reaches the target duration.

[0124] Furthermore, set slackness judgment standards, such as actual energy consumption is lower than a certain percentage of target energy consumption or actual exercise duration is lower than a certain proportion of target duration. Based on the comparison results, judge whether the user has not achieved the set goal. If the actual exercise data is significantly lower than the target data, it is considered that the user is slacking off. Generate specific reminder information based on the slackness detection results. For example, "Your exercise volume today has not reached the target, please pay attention to keep exercising." Slackness reminder information is pushed to users through mobile devices (such as mobile phones, smart watches, fitness mirrors, etc.) to provide timely feedback and incentives. Personalized suggestions and motivational words can be attached to the reminder information to help users re-stimulate their exercise motivation.

[0125] By comparing actual exercise data with preset target data, the system can monitor the user's training progress in real time and provide timely feedback. Slack reminders can motivate users to maintain their exercise habits and avoid affecting training results due to slackness. Through regular comparisons and reminders, users can better adhere to their training plans and achieve their fitness goals more effectively. The personalized suggestions and motivational words provided by the system can enhance the user's sense of participation and training motivation and improve the user experience. The continuous monitoring and feedback mechanism helps the system to continuously optimize the training plan and reminder mechanism and improve the overall intelligence level of fitness guidance. The system records actual exercise data (including total energy consumption and exercise duration) and compares it with the preset target data to determine whether the user is slacking. If slackness is detected, the system will generate and send a reminder message to the user's device to provide timely feedback and personalized suggestions. This process helps users adhere to their training plans, improve training results and user experience, and promote users to effectively achieve their fitness goals through timely monitoring and motivation.

[0126] This embodiment obtains fitness continuous frames by decoding and extracting frames of user fitness videos; performs human key point detection on each frame in the fitness continuous frames to obtain key point information; generates a human dynamic model based on the key point information; performs posture recognition on the human dynamic model based on a preset posture recognition model to obtain a motion posture set; if there is an erroneous motion posture in the motion posture set, generates correction data based on the standard posture corresponding to the erroneous motion posture in the preset standard posture library. This embodiment can finely analyze the movement of each moment to ensure the accuracy of the analysis by decoding and extracting frames of user fitness videos; uses human key point detection technology to accurately locate the joint position and provide data support for generating a dynamic model; the dynamic model fully reflects the user's motion posture, so that the system can monitor the movement in all directions and find subtle errors and potential risks. The preset posture recognition model can automatically analyze the posture, feedback errors in real time, and improve training efficiency. Finally, the correction data generated by the standard posture library can scientifically and accurately guide users to correct erroneous movements and avoid sports injuries, thereby greatly improving the fitness effect and ensuring that users can conduct effective training under safe and scientific guidance.

[0127] The present application also provides an auxiliary fitness device, please refer to Figure 4 , Figure 4 This is a schematic diagram of the module structure of the auxiliary fitness device according to an embodiment of the present application, and the auxiliary fitness device includes:

[0128] The pre-processing module 401 is used to decode and extract frames from the user's fitness video to obtain fitness continuous frames;

[0129] A key point detection module 402 is used to detect key points of the human body in each of the continuous fitness frames to obtain key point information;

[0130] A dynamic model generation module 403, used to generate a human body dynamic model based on the key point information;

[0131] A posture recognition module 404 is used to perform posture recognition on the human body dynamic model based on a preset posture recognition model to obtain a motion posture set;

[0132] The target correction module 405 is used to generate correction data based on a standard posture corresponding to the erroneous motion posture in a preset standard posture library if there is an erroneous motion posture in the motion posture set.

[0133] The auxiliary fitness device provided in the embodiment of the present application adopts the auxiliary fitness method in the above embodiment to solve the technical problem of how to improve the fitness effect. Compared with the prior art, the beneficial effects of the auxiliary fitness device provided in the embodiment of the present application are the same as the beneficial effects of the auxiliary fitness method provided in the above embodiment, and other technical features in the auxiliary fitness device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0134] The present application provides an auxiliary fitness device, which includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the auxiliary fitness method in the above embodiment.

[0135] Reference below Figure 5 , which shows a schematic diagram of the structure of an auxiliary fitness device suitable for implementing the embodiment of the present application. The auxiliary fitness device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The auxiliary fitness equipment shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0136] like Figure 5As shown, the auxiliary fitness device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of the auxiliary fitness device are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the auxiliary fitness device to communicate with other devices wirelessly or wired to exchange data. Although the figure shows an auxiliary fitness device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0137] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0138] The auxiliary fitness equipment provided by the present application adopts the auxiliary fitness method in the above embodiment to solve the technical problem of how to improve the fitness effect. Compared with the prior art, the beneficial effects of the auxiliary fitness equipment provided by the present application are the same as the beneficial effects of the auxiliary fitness method provided by the above embodiment, and other technical features in the auxiliary fitness equipment are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0139] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0140] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0141] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the auxiliary fitness method in the above-mentioned embodiment.

[0142] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above. The computer-readable storage medium may be included in the auxiliary fitness device; or may exist independently without being assembled into the auxiliary fitness device.

[0143] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the auxiliary fitness device, the auxiliary fitness device: decodes and extracts frames of the user's fitness video to obtain fitness continuous frames; performs human key point detection on each frame of the fitness continuous frames to obtain key point information; generates a human dynamic model based on the key point information; performs posture recognition on the human dynamic model based on a preset posture recognition model to obtain a motion posture set; if there is an erroneous motion posture in the motion posture set, generates correction data based on a standard posture corresponding to the erroneous motion posture in a preset standard posture library. The computer program code for performing the operation of the present application can be written in one or more programming languages ​​or a combination thereof, and the programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or it can be connected to an external computer (for example, through the Internet using an Internet service provider).

[0144] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0145] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0146] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned auxiliary fitness method, and can solve the technical problem of how to improve the fitness effect. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the auxiliary fitness method provided in the above-mentioned embodiment, and will not be repeated here.

[0147] An embodiment of the present application provides a computer program product, including a computer program, which implements the steps of the above-mentioned auxiliary fitness method when executed by a processor.

[0148] The computer program product provided in this application can solve the technical problem of how to improve fitness effects. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiment of this application are the same as the beneficial effects of the auxiliary fitness method provided in the above embodiment, which will not be repeated here.

[0149] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent processing scope of the present application.

Claims

1. A method for assisting fitness, characterized in that: The method comprises: Decode and extract frames of user fitness videos to obtain continuous fitness frames; Performing human body key point detection on each of the continuous fitness frames to obtain key point information; Based on the key point information, a human body dynamic model is generated; Based on a preset posture recognition model, performing posture recognition on the human body dynamic model to obtain a motion posture set; If there is an erroneous motion gesture in the motion gesture set, correction data is generated based on a standard gesture corresponding to the erroneous motion gesture in a preset standard gesture library.

2. The method according to claim 1, characterized in that The step of generating a human body dynamic model based on the key point information comprises: Based on a preset depth estimation algorithm, converting the key point information into three-dimensional coordinate information; Based on the three-dimensional coordinate information, a human skeleton model is generated; The human body dynamic model is generated based on the key point information and the human body skeleton model.

3. The method according to claim 1, characterized in that The step of performing posture recognition on the human body dynamic model based on a preset posture recognition model to obtain a motion posture set includes: Determine the posture frame sequence according to the human body dynamic model; Extracting features from the posture frame sequence, and inputting the feature extraction results into the preset posture recognition model; The motion gesture set is generated based on the output result of the preset gesture recognition model.

4. The method according to claim 1, characterized in that After the step of detecting the key points of the human body in each of the continuous fitness frames to obtain the key point information, the method further includes: Determine the changes in the bone structure according to the angle changes and speeds of each joint in the key point information; Based on the changes in the bone structure and a preset dangerous action detection model, a dangerous level assessment is performed on the current action; Based on the evaluation results, corresponding warning information is generated and pushed to the user's device.

5. The method according to claim 1, characterized in that After the step of performing posture recognition on the human body dynamic model based on the preset posture recognition model to obtain a motion posture set, the method further includes: Based on a preset action classification model, classifying the motion postures in the motion posture set into multiple types; Based on the preset energy consumption model, determine the energy consumption per unit time of each type of motion posture; Determine the total energy consumption according to the type of each motion posture in the motion posture set and the corresponding motion duration; Adjust fitness plans and dietary recommendations based on the user's training goal data and the total energy consumption.

6. The method according to claim 5, characterized in that After the step of determining the energy consumption per unit time of each type of motion posture based on the preset energy consumption model, the method further includes: The total amount of energy consumption and the duration of the motion posture set are used as actual motion data; Comparing the actual exercise data with the preset target data, and judging whether the user is slacking off based on the comparison result; If yes, send a slack reminder message to the user's device.

7. An auxiliary fitness device, characterized in that: The device comprises: A preprocessing module is used to decode and extract frames from user fitness videos to obtain fitness continuous frames; A key point detection module, used to detect the key points of the human body in each of the continuous fitness frames to obtain key point information; A dynamic model generation module, used to generate a human body dynamic model based on the key point information; A posture recognition module, used to perform posture recognition on the human body dynamic model based on a preset posture recognition model to obtain a motion posture set; The target correction module is used to generate correction data based on a standard posture corresponding to the erroneous motion posture in a preset standard posture library if there is an erroneous motion posture in the motion posture set.

8. A computer device, characterized in that: The device comprises: a memory, a processor, and an auxiliary fitness program stored in the memory and executable on the processor, wherein the auxiliary fitness program is configured to implement the steps of the auxiliary fitness method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores an auxiliary fitness program, and when the auxiliary fitness program is executed by the processor, the steps of the auxiliary fitness method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the auxiliary fitness method according to any one of claims 1 to 6 are implemented.

Citation Information

Cited By

  • Posture correction training method and device applied to human intelligence manufacturing

    CN119366911A

  • A posture correction training method and device applied to human-centered intelligent manufacturing

    CN119366911B

  • Monitoring method and system based on health data in exercise program of sporter

    CN119694575A

  • Personalized motion scheme generation and adjustment method based on motion offset degree

    CN120048427A

  • Complex posture behavior analysis method based on machine learning

    CN121167252A