Fitness posture joint monitoring method based on multi-dimensional vision and sports wearable equipment

Abstract

The invention discloses a fitness posture joint monitoring method based on multi-dimensional vision and a sports wearable device. The method comprises the following steps: acquiring user sports information through a multi-dimensional movable vision fitness monitoring device; obtaining a motion signal and a physiological signal through wearable motion detection equipment; training a preset neural network model based on the user motion information to obtain a motion capture model; inputting the user motion information obtained in real time into the motion capture model, judging whether the user motion meets the standard or not, and performing real-time feedback according to the judgment result; constructing an exercise health management prediction model, and training the exercise health management prediction model based on the exercise signals and the physiological signals; and generating a health management prediction report of the user through the trained model. Existing fitness equipment is optimized, a portable design is adopted, combined monitoring of a contact sensor and non-contact equipment is combined, and comprehensive monitoring and guidance are provided for a user from a multi-dimensional perspective.

Description

Technical Field

[0001] The present invention belongs to the technical field of sports health monitoring, and particularly relates to a fitness posture joint monitoring method based on multi-dimensional vision and wearable sports devices. Background Art

[0002] In recent years, the public's demand for scientific fitness has been increasing. However, the current fitness environment faces many challenges. On the one hand, the number of professional fitness instructors is limited, making it difficult to meet the needs of the vast number of fitness enthusiasts, and the professional levels of the instructors vary, so the quality of fitness guidance is difficult to guarantee. On the other hand, the costs of fitness venues and professional guidance are high, discouraging many fitness enthusiasts. Against this background, most self-exercising fitness enthusiasts are prone to unconsciously repeat incorrect movements due to the lack of correct movement technique guidance, resulting in sports injuries.

[0003] Under the current technical conditions, although some fitness devices are equipped with simple sensors, the data provided by these sensors is limited and cannot comprehensively monitor and analyze the user's movement actions. Usually, it is limited to a certain part, or only monitors some health indicators of the human body. There are few devices that can synchronously monitor the whole-body movement state and provide real-time feedback, and cannot provide comprehensive and systematic guidance for users, with great limitations. Moreover, these devices usually need to be used in a specific environment, restricting the breadth of their applications. And some applications that provide movement guidance through video analysis also have problems such as insufficient accuracy and poor real-time performance in actual applications.

[0004] Therefore, there is an urgent need for a more comprehensive, accurate, and convenient solution to meet the growing demand for scientific fitness. Summary of the Invention

[0005] To solve the above technical problems, the present invention proposes a fitness posture joint monitoring method based on multi-dimensional vision and wearable sports devices to solve the problems existing in the above-mentioned prior art.

[0006] To achieve the above object, the present invention provides a fitness posture joint monitoring method based on multi-dimensional vision and wearable sports devices, including:

[0007] Integrated deployment of a multi-dimensional movable vision fitness monitoring device and a wearable motion detection device;

[0008] Obtaining user movement information through the multi-dimensional movable vision fitness monitoring device; obtaining motion signals and physiological signals through the wearable motion detection device;

[0009] Train a preset neural network model with user movement information to obtain an action capture model; input the real-time obtained user movement information into the action capture model to determine whether the user's actions meet the standards, and give real-time feedback according to the judgment results;

[0010] Construct a sports health management prediction model, and train the sports health management prediction model based on the movement signals and physiological signals; generate a health management prediction report for the user through the sports health management prediction model.

[0011] Optionally, the multi-dimensional movable visual fitness monitoring device includes four portable camera devices, and adjacent camera devices are connected by retractable elastic ropes; the camera device includes a main body and an additional fixing device; the additional fixing device includes a ring outside the main body and a fixing device, and there are small holes on the ring.

[0012] Optionally, the process of training a preset neural network model with user movement information to obtain an action capture model includes:

[0013] Obtain user movement information during the historical training process through a multi-dimensional movable visual fitness monitoring device, where the user movement information includes safe action samples and incorrect action samples; use the safe action samples as the first samples and the incorrect action samples as the second samples to train the preset neural network model to obtain an action capture model. Among them, extract the user action key point information from the user movement information and input it into the action capture model.

[0014] Optionally, the process of inputting the real-time obtained user movement information into the action capture model to determine whether the user's actions meet the standards and giving real-time feedback according to the judgment results includes:

[0015] Obtain user status data through a multi-dimensional movable visual fitness monitoring device, judge whether the user's physiological status meets the standards according to the user status data within a preset time period. If it does not meet the standards, generate a warning signal. If it meets the standards, input the real-time obtained user movement information into the action capture model to judge whether the user's actions meet the standards. If they do not meet the standards, generate a reminder signal.

[0016] Optionally, the movement signals include electromyogram signals and tension signals, and the physiological signals include electrocardiogram signals and acceleration energy.

[0017] Optionally, the sports health management prediction model includes an action tracking network and a health status prediction network. Train the action tracking network based on the movement signals and train the health status prediction network based on the physiological signals.

[0018] Optionally, the process of training the action tracking network based on the movement signals includes:

[0019] After filtering and method processing the motion signal, it is converted into digital information. Based on the digital signal, the rotation angle of the user's motion joints is calculated. Based on the preset kinematic model and the rotation angle of the motion key, the spatial pose coordinates of the current position points of each motion joint relative to the initial point are calculated to obtain motion pose information. The action tracking network is trained based on the user motion information and the corresponding standard judgment results, the motion pose information, the motion signal, the user motion information marking data, and the motion signal marking data.

[0020] Optionally, the process of training the health status prediction network based on physiological signals includes:

[0021] Use multiple base learners for training to obtain multiple wavelet threshold functions with different accuracies;

[0022] Use the preset particle swarm optimization algorithm to calculate the optimal combination weights of the parameters of the wavelet threshold function corresponding to each base learner to generate an adaptive wavelet threshold function;

[0023] Substitute the physiological signal into the adaptive wavelet threshold function and perform threshold processing using the unified threshold method to obtain the wavelet coefficients after threshold processing; reconstruct the wavelet coefficients to obtain the reconstructed physiological signal;

[0024] Train the health status prediction network based on the reconstructed physiological signal and the physiological signal marking data.

[0025] The present invention also provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the computer program to implement the steps of the above method.

[0026] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0027] Compared with the prior art, the present invention has the following advantages and technical effects:

[0028] The present invention first integrally deploys a multi-dimensional movable visual fitness monitoring device and a wearable motion detection device; then, obtains user motion information through the multi-dimensional movable visual fitness monitoring device, and obtains motion signals and physiological signals through the wearable motion detection device; next, trains a preset neural network model through the user motion information to obtain an action capture model, inputs the real-time obtained user motion information into the action capture model to determine whether the user's actions meet the standards, and gives real-time feedback according to the judgment results; finally, constructs a motion health management prediction model, trains the motion health management prediction model based on the motion signals and physiological signals, and generates a health management prediction report for the user through the motion health management prediction model. By optimizing the existing fitness equipment, adopting a portable design, and combining the joint monitoring of contact sensors and non-contact devices, the present invention provides comprehensive monitoring and guidance for fitness enthusiasts and the public from a multi-dimensional perspective. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0030] Figure 1 It is a schematic diagram of the method according to an embodiment of the present invention;

[0031] Figure 2 It is a schematic diagram of the overall folding perspective of the camera according to an embodiment of the present invention;

[0032] Figure 3 It is a schematic diagram of the overall separated perspective of the camera according to an embodiment of the present invention;

[0033] Figure 4 It is an enlarged top view of a single camera according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0035] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0036] Embodiment 1

[0037] At present, the fitness equipment on the market is mainly divided into aerobic exercise equipment, strength training equipment, multi-functional fitness equipment, auxiliary fitness equipment, etc. This type of equipment is designed to provide trainers with targeted training methods to meet the training needs of different trainers. However, in the actual independent training process, fitness enthusiasts often lack professional guidance, resulting in inadequate training results and even injuries.

[0038] In addition, the concept of exercise is not limited to fitness and bodybuilding, and the extension of the concept of exercise is expanded. Office workers sitting and typing for a long time, students writing at a desk for a long time, etc. are all a kind of exercise. Incorrect sitting and standing postures may cause health risks. Correcting these bad health behavior habits in time through the present invention will bring more benefits to the health of the public.

[0039] In recent years, thanks to the popularity of sensors and intelligent recognition technology, various types of sports monitoring equipment have received more and more attention. However, existing sports monitoring equipment is usually limited to a certain part, or only monitors some health indicators of the human body. Few can simultaneously monitor the whole body's exercise status and provide real-time feedback, and cannot provide users with comprehensive and systematic guidance, which has great limitations.

[0040] The present invention aims to enhance the scientific nature of the independent training of fitness enthusiasts. By optimizing existing fitness equipment, adopting a portable design, and combining the joint monitoring of contact sensors and non-contact devices, it provides comprehensive monitoring and guidance for fitness enthusiasts and the general public from a multi-dimensional perspective.

[0041] like Figures 1-4 As shown, this embodiment provides a fitness posture joint monitoring method based on multi-dimensional vision and sports wearable devices, including:

[0042] Integrated deployment of multi-dimensional mobile visual fitness monitoring equipment and wearable motion detection equipment;

[0043] The user's motion information is obtained through multi-dimensional movable visual fitness monitoring equipment; motion signals and physiological signals are obtained through wearable motion detection equipment; motion signals include electromyography signals and tension signals, and physiological signals include electrocardiogram signals and acceleration energy.

[0044] Furthermore, the multi-dimensional movable visual fitness monitoring equipment includes four portable camera devices, and adjacent camera devices are connected by retractable elastic ropes; the camera device includes a main body and an additional fixing device; the additional fixing device includes a circular ring on the outside of the main body and a fixing device, and a small hole is provided on the circular ring.

[0045] Specifically, the multi-dimensional movable visual fitness monitoring device consists of four lightweight and portable cameras. When not in use, it can be folded up. The four cameras can be folded into a cylindrical shape by the magnetic attraction part on the outermost circle. Each camera is connected to its adjacent camera by a retractable elastic cord. When in use, the elastic cord is extended to freely create one's own fitness space. Each camera device is composed of a main body part and an additional fixing device part. The main body part is in the shape of a small cylinder. The innermost small circle is the camera, and there are three small holes on the ring close to the camera, which facilitate the detachable replacement of different plugins (magnetic attraction, suction cup, mechanical fixing device). Different plugins are used to adapt to different environments, and the specific operation can be selected and assembled by the user himself.

[0046] Train a preset neural network model through the user's motion information to obtain an action capture model; input the user's motion information obtained in real time into the action capture model, judge whether the user's actions meet the standards, and give real-time feedback according to the judgment results;

[0047] Furthermore, the process of training a preset neural network model through the user's motion information to obtain an action capture model includes:

[0048] Obtain the user's motion information during the historical training process through the multi-dimensional movable visual fitness monitoring device. The user's motion information includes safe action samples and incorrect action samples; use the safe action samples as the first samples and the incorrect action samples as the second samples to train the preset neural network model to obtain an action capture model. Among them, extract the user's action key point information from the user's motion information and input it into the action capture model.

[0049] Furthermore, the process of inputting the user's motion information obtained in real time into the action capture model, judging whether the user's actions meet the standards, and giving real-time feedback according to the judgment results includes:

[0050] Obtain the user's status data through the multi-dimensional movable visual fitness monitoring device, judge whether the user's physiological status meets the standards according to the user's status data within a preset time period. If it does not meet the standards, generate a warning signal. If it meets the standards, input the user's motion information obtained in real time into the action capture model, judge whether the user's actions meet the standards, and generate a reminder signal if they do not meet the standards.

[0051] Specifically, the multi-dimensional movable visual fitness monitoring device performs the following operations:

[0052] (1) Action sample collection and key point extraction

[0053] Action sample collection: Four cameras are used to comprehensively capture the user's action samples during historical training, including safe action samples and incorrect action samples. The arrangement of the cameras ensures that all angles of the user's movement can be covered, thereby obtaining comprehensive action information.

[0054] Key point extraction: The safe action samples and incorrect action samples are processed separately to extract the three-dimensional data set of the user's action key points. These key points include, but are not limited to, joint positions, body postures, etc., which can accurately reflect the characteristics of the user's actions.

[0055] (2) Action capture model training

[0056] Sample input and model training: The safe action samples are used as the first samples, and the incorrect action samples are used as the second samples and input into a preset neural network model for training. By comparing and analyzing the characteristics of safe and incorrect action samples, the model can learn the key feature differences in different action states and update the model parameters accordingly, finally generating an action capture model.

[0057] (3) Evaluation of the current user's posture state

[0058] Real-time data input and evaluation: The three-dimensional data set corresponding to the posture state data of the current training user is input into the trained action capture model. By comparing and analyzing the characteristics of the user's current posture and safe action samples, the model can judge in real time whether the user's fitness posture actions meet the safety and standard requirements and output the evaluation results.

[0059] (4) Data collection and monitoring

[0060] Data collection end design: The data collection end includes the following modules:

[0061] Data monitoring module: Used to monitor the user's state data information in real time, including physiological data (such as heart rate, blood pressure, etc.) and motion data (such as action amplitude, speed, etc.).

[0062] Data analysis module: Analyze the user's physiological state data within a preset time period to judge whether it meets the preset safety and health standards.

[0063] Feedback information generation: When the user's physiological state data does not meet the requirements, the system generates a warning signal and issues a warning to the user through the terminal output device (such as mobile phone, tablet, etc.), prompting the user to pause the exercise or take corresponding measures; if it meets the requirements, feedback that the action is correct.

[0064] On the one hand, it can enable fitness enthusiasts to recognize what the correct posture is and strengthen the correct cognition; on the other hand, the recognition of the correct posture can improve the confidence of fitness enthusiasts and obtain positive feedback.

[0065] Analysis of postural state data: When the user's physiological state data meets the requirements, further analyze the user's postural state data to determine whether it conforms to the standard fitness posture.

[0066] Data collection: If the user's postural state conforms to the standard, collect the current data as the basis for subsequent model optimization and the user's exercise history record.

[0067] Reminder signal generation: If the user's postural state does not conform to the standard, generate a reminder signal and send a postural adjustment reminder to the user through the terminal output device to guide the user to correct the posture.

[0068] Build a sports health management prediction model and train the sports health management prediction model based on sports signals and physiological signals; generate a health management prediction report for the user through the sports health management prediction model.

[0069] Furthermore, the sports health management prediction model includes an action tracking network and a health status prediction network. The action tracking network is trained based on sports signals, and the health status prediction network is trained based on physiological signals.

[0070] Furthermore, the process of training the action tracking network based on sports signals includes:

[0071] After filtering and method processing the sports signals, convert them into digital information, calculate the rotation angles of the user's movement joints based on the digital signals, calculate the spatial pose coordinates of the current position points of each movement joint relative to the initial point based on the preset kinematic model and the key rotation angles of the movement, and obtain the movement posture information; train the action tracking network based on the user's movement information and the corresponding standard judgment results, movement posture information, sports signals, user movement information marking data, and sports signal marking data.

[0072] Furthermore, the process of training the health status prediction network based on physiological signals includes:

[0073] Use multiple base learners for training to obtain multiple wavelet threshold functions with different accuracies;

[0074] Use the preset particle swarm algorithm to calculate the optimal combination weights of the parameters of the wavelet threshold functions corresponding to each base learner to generate an adaptive wavelet threshold function;

[0075] Substitute the physiological signals into the adaptive wavelet threshold function, perform threshold processing using the unified threshold method to obtain the wavelet coefficients after threshold processing; reconstruct the wavelet coefficients to obtain the reconstructed physiological signals;

[0076] Train the health status prediction network based on the reconstructed physiological signals and the physiological signal marking data.

[0077] After fitness enthusiasts wear the device of the present invention, the device can collect and process various motion and physiological signals in real time. Specifically, it includes:

[0078] (1) The first detection module: It is used to detect the electromyogram signals generated by the user's skeletal muscles. This module includes electromyogram electrodes and a first band-pass amplifier. Through filtering and amplification processing, the electromyogram signals are converted into first digital signals.

[0079] The second detection module: It is used to detect the tension signals on the user's skin surface. This module includes skin surface tension strain gauges and a second band-pass amplifier. After filtering and amplification processing, the tension signals are converted into second digital signals.

[0080] (2) Motion posture information processing

[0081] The first calculation module: Receives the digital signals output by the first detection module and the second detection module, and calculates the rotation angles of the user's motion joints through a first preset algorithm. Further, in combination with the pre-established kinematic model, it calculates the spatial pose coordinates of the current position points of each motion joint relative to the initial point, thereby generating the user's motion posture information.

[0082] The generation of the first control signal: Based on the above motion posture information, a first control signal is generated through a second preset algorithm, which is used to provide real-time feedback and guide the adjustment of the user's motion posture.

[0083] Monitoring of motion parameters and data transmission

[0084] Collection of motion parameters: The device can obtain various motion parameters during the user's exercise in real time, including electrocardiogram signals and acceleration energy, etc.

[0085] Data transmission: The local wireless communication node receives the motion monitoring data sent by the device and transmits it to the user's terminal device (such as a mobile phone or a computer) through the wireless network, so that the user can view and analyze the motion data at any time.

[0086] (4) Health management prediction model

[0087] Model construction: An initial health management prediction model is constructed, which includes a primary motion tracking data analysis network and a primary health status prediction network. The output ends of the two are connected to the input ends to form a complete prediction system.

[0088] (5) Generation of user motion monitoring and health management reports

[0089] Generation of motion monitoring data: The terminal device generates user motion monitoring data based on the collected motion parameters and the preset target motion parameters, and marks them.

[0090] Health management prediction report generation: Input the user's exercise monitoring data into the trained health management prediction model to generate a health management prediction report after the user's exercise.

[0091] Optimization of target exercise parameters: Optimize the preset target exercise parameters according to the health management prediction report to generate optimized target exercise parameters.

[0092] Exercise monitoring report generation: Combine the optimized target exercise parameters and the actual exercise parameters to generate a user exercise monitoring report, providing personalized exercise guidance and health management suggestions for the user. After each use by the user, a summary report is generated after data analysis and presented on the terminal device, such as the frequency of errors, what the errors are respectively, the correction effect after prompting, what improvement methods and directions are available in the future, using motivating and encouraging words.

[0093] This embodiment also provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the steps of the above method.

[0094] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0095] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices, characterized in that It includes the following steps: Integrate and deploy a multi-dimensional movable visual fitness monitoring device and a wearable motion detection device; Obtain user motion information through the multi-dimensional movable visual fitness monitoring device; Obtain motion signals and physiological signals through the wearable motion detection device; Train a preset neural network model with the user motion information to obtain an action capture model; input the real-time obtained user motion information into the action capture model to determine whether the user's actions meet the standards, and give real-time feedback according to the judgment results; Construct a sports health management prediction model and train the sports health management prediction model based on the motion signals and physiological signals; generate a health management prediction report for the user through the sports health management prediction model.

2. The fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices according to claim 1, wherein The multi-dimensional movable visual fitness monitoring device includes four portable camera devices, and adjacent camera devices are connected by retractable elastic ropes; the camera device includes a body and an additional fixing device; the additional fixing device includes a ring outside the body and a fixing device, and small holes are provided on the ring.

3. The fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices according to claim 1, wherein The process of training a preset neural network model with the user motion information to obtain an action capture model includes: Obtain the user motion information during the historical training process through the multi-dimensional movable visual fitness monitoring device, and the user motion information includes safe action samples and incorrect action samples; use the safe action samples as the first samples and the incorrect action samples as the second samples to train the preset neural network model to obtain an action capture model, wherein, extract the user action key point information from the user motion information and input it into the action capture model.

4. The fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices according to claim 1, wherein The process of inputting the real-time obtained user motion information into the action capture model to determine whether the user's actions meet the standards and giving real-time feedback according to the judgment results includes: Obtain the user status data through the multi-dimensional movable visual fitness monitoring device, and determine whether the user's physiological status meets the standards according to the user status data within a preset time period. If it does not meet the standards, generate a warning signal. If it meets the standards, input the real-time obtained user motion information into the action capture model to determine whether the user's actions meet the standards. If not, generate a reminder signal.

5. The fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices according to claim 1, wherein The motion signals include electromyography signals and tension signals, and the physiological signals include electrocardiogram signals and acceleration energy.

6. The fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices according to claim 1, wherein The motion health management prediction model includes an action tracking network and a health status prediction network. The action tracking network is trained based on the motion signals, and the health status prediction network is trained based on the physiological signals.

7. The fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices according to claim 6, wherein The process of training the action tracking network based on the motion signals includes: After filtering and method processing the motion signals, they are converted into digital information. Based on the digital signals, the rotation angles of the user's motion joints are calculated. Based on the preset kinematic model and the key rotation angles of the motion, the spatial pose coordinates of the current position points of each motion joint relative to the initial point are calculated to obtain motion posture information. The action tracking network is trained based on the user's motion information and the corresponding standard judgment results, the motion posture information, the motion signals, the user's motion information marking data, and the motion signal marking data.

8. The fitness posture joint monitoring method based on multi-dimensional vision and motion wearable devices according to claim 6, wherein The process of training the health status prediction network based on the physiological signals includes: Multiple base learners are used for training to obtain multiple wavelet threshold functions with different accuracies; The particle swarm optimization algorithm is used to calculate the optimal combination weights of the parameters of the wavelet threshold functions corresponding to each base learner to generate an adaptive wavelet threshold function; The physiological signals are substituted into the adaptive wavelet threshold function, and threshold processing is performed using the unified threshold method to obtain the wavelet coefficients after threshold processing. The wavelet coefficients are reconstructed to obtain the reconstructed physiological signals; The health status prediction network is trained based on the reconstructed physiological signals and the physiological signal marking data.

9. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1-8.

Images