A smart garment for fitness motion capture and body posture reconstruction

Abstract

This invention provides a smart garment for capturing fitness movements and reconstructing body postures, comprising a wearable device body, a data collection system, and a data processing system. The wearable device body includes a tight-fitting sportswear, triboelectric strain sensing fibers, flexible conductive circuitry, and a packaging shell. The data collection system includes a data acquisition chip and a power module. The data processing system includes a display terminal that executes a comprehensive algorithm, which includes a deep learning algorithm, a multi-dimensional matching calibration algorithm, a feature extraction algorithm, and a body posture reconstruction algorithm. The smart garment can achieve precise full-space fitness movement data capture and body posture reconstruction. By collecting motion signals during fitness activities, identifying the name of the fitness activity, calculating joint bending angle data and comparing it with standard data, the garment can display the movement posture configuration in real time in a three-dimensional visualization simulation environment and provide real-time fitness feedback suggestions.

Description

Technical Field

[0001] This invention relates to the field of smart wearable device technology, and in particular to a smart garment for fitness motion capture and body posture reconstruction. Background Technology

[0002] With the development of technology and the improvement of people's living standards, people are increasingly aware of the importance of home fitness. Proper form is crucial during home workouts; incorrect posture can lead to joint injuries due to improper joint movement. However, people who exercise at home often lack professional guidance and cannot verify the correctness of their postures from a third-party perspective, making it difficult to collect and analyze exercise data and make appropriate corrections.

[0003] The hardware devices used for motion capture in commercially available smart fitness products can be divided into optical cameras and external inertial devices. However, optical motion capture technology is limited by spatial constraints and privacy risks, while motion capture devices based on inertial sensors suffer from inherent error accumulation and low comfort levels. Therefore, smart clothing based on wearable flexible sensors has broad application prospects in virtual reality, entertainment, and medical fields due to its advantages such as convenience, comfort, high-precision real-time monitoring, lack of spatial and exercise type limitations, and multi-functional integration.

[0004] However, existing research on smart clothing based on wearable flexible sensors still has some problems to solve, including: 1. High-precision recognition of similar movements in smart clothing. 2. Good versatility in smart clothing. 3. Accurate reconstruction of body postures for fitness. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a smart garment for fitness motion capture and body posture reconstruction.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] A smart garment for fitness motion capture and body posture reconstruction includes a wearable device, a data collection system, and a data processing system, wherein...

[0008] The wearable device body includes a tight-fitting sportswear, triboelectric strain sensing fibers, flexible conductive circuits, and an encapsulation shell. The triboelectric strain sensing fibers and flexible conductive circuits are integrated into the tight-fitting sportswear using textile technology, and the encapsulation shell is disposed on the tight-fitting sportswear.

[0009] The data collection system includes a data acquisition chip and a power module. The data acquisition chip and the power module are connected by circuits and disposed inside the encapsulation shell. The data acquisition chip is connected to the triboelectric strain sensing fiber through a flexible conductive circuit. The data acquisition chip is used to collect data from the triboelectric strain sensing fiber.

[0010] The data processing system includes a display terminal that executes a comprehensive algorithm, which includes a deep learning algorithm, a multi-dimensional matching calibration algorithm, a feature extraction algorithm, and a body posture reconstruction algorithm. The display terminal is connected to a data acquisition chip and receives data from the triboelectric strain sensing fiber acquired by the chip. After data analysis by the comprehensive algorithm within its application, the system displays the body posture configuration and provides real-time fitness feedback suggestions. Specifically, the deep learning algorithm identifies the movement name based on the output voltage signal of the acquired triboelectric strain sensing fiber, then calculates the joint angle data from the output voltage signal using the multi-dimensional matching calibration algorithm, then determines the macroscopic center of gravity, position, and orientation of the virtual character model based on the voltage signal using the feature extraction algorithm, and finally controls the bending angles and macroscopic state changes of each joint of the virtual character based on the previous calculation results using the body posture reconstruction algorithm. This achieves high-precision body posture reconstruction for fitness exercises in a three-dimensional visualized simulation environment and provides real-time fitness feedback suggestions.

[0011] Preferably, the aforementioned smart clothing for fitness motion capture and body posture reconstruction includes a long-sleeved, long-legged tight-fitting sportswear, quick-drying clothing, yoga training clothing, or kung fu training clothing in regular clothing sizes.

[0012] Preferably, in the above-mentioned smart clothing for fitness motion capture and body posture reconstruction, the triboelectric strain sensing fibers are integrated into the corresponding positions of the back, shoulder joint, scapula, elbow joint and / or knee joint of the tight-fitting sportswear. This not only enables accurate measurement of the user's joint angles, but also does not cause skin allergies.

[0013] Preferably, in the above-mentioned smart clothing for fitness motion capture and body posture reconstruction, the triboelectric strain sensing fiber is composed of a flexible conductive circuit, an encapsulation layer, a dielectric layer, and a working electrode layer. The dielectric layer completely encapsulates the working electrode layer. The flexible conductive circuit is located at both ends of the dielectric layer. The encapsulation layer is located at the junction of the flexible conductive circuit and the dielectric layer. The encapsulation layer is made of polydimethylsiloxane, Ecoflex, or UV-curable adhesive. The dielectric layer is an elastic conduit. The working electrode layer is made of liquid metal.

[0014] Preferably, in the above-mentioned smart clothing for motion capture and body posture reconstruction, the liquid metal is mercury, francium, cesium, gallium, gallium-indium alloy, gallium-tin alloy, gallium-zinc alloy, gallium-indium-tin alloy, gallium-indium-zinc alloy, or liquid sodium-potassium alloy; the elastic conduit is a silicone tube, a polyurethane tube, or a polyethylene tube.

[0015] Preferably, in the above-mentioned smart clothing for fitness motion capture and body posture reconstruction, the method for preparing the triboelectric strain sensing fiber is to inject liquid metal into an elastic conduit through an injection pump until the cavity is filled, insert flexible conductive lines from both ends of the elastic conduit, and then seal the ports of the elastic conduit with encapsulation material to prevent liquid metal leakage.

[0016] Preferably, in the above-mentioned smart clothing for fitness motion capture and body posture reconstruction, the triboelectric strain sensing fiber and flexible conductive circuit are integrated into the tight-fitting sportswear by means of weaving, knitting, weaving, embroidery or sewing. The material of the flexible conductive circuit is silicone wire, PVC wire, insulated copper wire, insulated aluminum wire, polyester film wire, Teflon wire, nano silver wire or fluoroplastic wire.

[0017] Preferably, the smart clothing used for motion capture and body posture reconstruction integrates patterns including meanders, zigzags, waves, lines, or serrations.

[0018] Preferably, in the aforementioned smart clothing for fitness motion capture and body posture reconstruction, the data acquisition chip includes a data processor with a detachable data interface and a wireless signal transmission module. The data processor can effectively receive data measured by triboelectric strain sensing fibers and transmit it to the data processing system through the wireless signal transmission module. At the same time, the data processor and the data interface are detachable, which facilitates the disassembly, maintenance and repair of the data processor.

[0019] Preferably, in the aforementioned smart clothing for capturing fitness movements and reconstructing body posture, the interface of the flexible conductive circuit is a pluggable interface, used to connect the data processor of the triboelectric strain sensing fiber and the data acquisition chip, and is integrated into the surface of the tight-fitting sportswear for data transmission.

[0020] Preferably, in the aforementioned smart clothing for motion capture and body posture reconstruction, the integrated algorithm is installed in the display terminal via an application.

[0021] Preferably, in the aforementioned smart clothing for motion capture and body posture reconstruction, the data acquisition chip is connected to the application signal within the display terminal via Bluetooth or a wireless network.

[0022] Preferably, in the above-mentioned smart clothing for capturing fitness movements and reconstructing body postures, the application is an APP software or a mini-program used to display the synchronized movement of a virtual character model during fitness.

[0023] Preferably, in the above-mentioned smart clothing for motion capture and body posture reconstruction, the display terminal is an electronic display device connected via wireless transmission technology, including a mobile phone, computer, display screen, projector, television, or VR glasses.

[0024] Preferably, in the above-mentioned smart clothing for fitness motion capture and body posture reconstruction, the integrated algorithm includes:

[0025] The deep learning algorithm is developed based on a convolutional neural network model, and its main function is to accurately identify the names of fitness exercises.

[0026] The multi-dimensional matching calibration algorithm is developed based on a locally weighted linear regression model. Its main function is to accurately calculate the joint bending angle while ignoring differences in user body shape.

[0027] The feature extraction algorithm is developed based on a machine learning model, and its main function is to determine the macroscopic center of gravity, position, and orientation of the virtual character model.

[0028] The body posture reconstruction algorithm is developed using the C# compiler language through the Unity-3D platform. Its main function is to control the changes in the joint angles and macroscopic state of the virtual character, compare the calculated angle of joint bending with the standard bending angle to determine whether the fitness movement is standard, and display the results on the display terminal.

[0029] Preferably, in the aforementioned smart clothing for motion capture and body posture reconstruction, the deep learning algorithm includes the following steps:

[0030] (1) Data preparation: Preprocess the collected raw data, including outlier removal, missing data filling, filtering, normalization and data segmentation, and integrate the results into a reliable dataset;

[0031] (2) Signal similarity analysis: Spearman correlation analysis was used to analyze the signal similarity of the collected body postures, calculate the correlation coefficient of any two body posture signals, and statistically analyze the correlation coefficient distribution matrix.

[0032] (3) Training the 1D-CNN deep learning algorithm model: 80% of the data samples in the reliable dataset were used for training for a total of 300 training epochs, and 20% were used to test the accuracy of the 1D-CNN deep learning algorithm model. The 1D-CNN model contains 4 kernels, 64 filters, and 4 convolutional layers;

[0033] (4) Evaluation of 1D-CNN deep learning algorithm model: The efficiency of 1D-CNN model is evaluated by learning accuracy and learning rate, and the reliability of body pose recognition is evaluated by confusion matrix.

[0034] Preferably, in the above-mentioned smart clothing for fitness motion capture and body posture reconstruction, the multi-dimensional matching calibration algorithm includes the following steps:

[0035] (1) Basic data preparation: Collect the joint dimensions and height information of several volunteers, and collect body posture signals according to the preset 24 body postures;

[0036] (2) Database establishment: Use support vector regression model to statistically analyze the distribution relationship between joint size and output signal for a specified joint bending angle, and use ordinary regression analysis method to statistically analyze the distribution relationship between joint bending angle and output voltage for a specified joint name. Store the above statistical results in the database.

[0037] (3) Self-calibration: Volunteers collect body posture signals according to the five preset calibration actions. Then, the Spearman correlation analysis method is used to match the distribution relationship model of the size of each joint and the output signal in the database. Next, the two-dimensional database is used to search for the distribution relationship model of the joint bending angle and the output voltage on both sides of the joint size. Finally, the local weighted linear regression analysis is used to achieve accurate correction of the individual regression relationship between the joint angle and the output voltage.

[0038] (4) Joint angle calculation: The real-time joint bending angle is obtained by calculating the output signal of each joint in sequence through the corrected individual regression relationship.

[0039] Preferably, in the above-mentioned smart clothing for motion capture and body posture reconstruction, the feature extraction algorithm includes the following steps:

[0040] (1) Data preparation: Isolate the voltage signals of the knee and back joints in the body posture signal;

[0041] (2) Feature extraction: The decision tree algorithm is used to analyze the temporal features of the knee and back joints and extract features including waveform, trough time, peak time, trough value, peak value, trough amount, number of peaks, high pressure duration, mean, variance, standard deviation, root mean square and kurtosis.

[0042] (3) Spatial element assignment: The center of gravity, position and orientation of the spatial elements of the virtual model are assigned based on the feature extraction results and the statistical height information. The data for assignment are calculated from height based on the behavior analysis statistical formula, calculated from height and position deviation based on the spatial geometric model, and estimated from gait analysis statistical results.

[0043] Preferably, in the above-mentioned smart clothing for fitness motion capture and body posture reconstruction, the body posture reconstruction algorithm includes the following steps:

[0044] (1) Virtual character model drawing: Fine-tune the preset virtual character model size, skeleton and joint controller position according to the volunteer's height;

[0045] (2) Control program writing: The real-time joint angle data calculated by the multi-dimensional matching calibration algorithm and the center of gravity, position and orientation deviation data calculated by the feature extraction algorithm are sent to the Unity-3D platform via Socket communication, and compiled by the C# compiler into joint rotation commands and bone movement commands based on Euler angles;

[0046] (3) Virtual character body posture reconstruction: The skeleton and joint controller executes control commands to drive the virtual character model to change posture.

[0047] Beneficial effects:

[0048] The aforementioned smart clothing for fitness motion capture and body posture reconstruction can achieve precise full-space fitness motion data capture and body posture reconstruction. By collecting motion signals during fitness activities, identifying the name of the fitness activity, calculating joint bending angle data and comparing it with standard data, the posture configuration of the body fitness activities is displayed in real time in a three-dimensional visualization simulation environment, and fitness feedback suggestions are provided in real time, realizing standardized and efficient home fitness.

[0049] When exercising with the smart clothing, the strain magnitude of the nine joints throughout the body is sensed by triboelectric strain sensing fibers and transmitted to a display terminal via a portable data acquisition chip. Deep learning algorithms improve the accuracy of recognizing similar fitness exercise names. A multi-dimensional matching calibration algorithm calculates the actual bending angle of the nine joints and compares it with standard joint bending angles to determine whether the exercise is standard. Feature extraction algorithms determine the macroscopic center of gravity, position, and orientation of the virtual character model. A body posture reconstruction algorithm reconstructs the body posture of the virtual character model using real-time joint angle data and macroscopic state data. The display terminal synchronously displays the fitness exercises of the virtual character model and provides fitness feedback suggestions. This allows users to achieve immersive and interactive exercise. Furthermore, the comprehensive algorithm eliminates test data errors caused by differences in user body shape, improving the versatility of the smart clothing. Attached Figure Description

[0050] Figure 1 It is the triboelectric strain sensing fiber described in this invention.

[0051] Figure 2This is a front structural diagram of the smart clothing for fitness motion capture and body posture reconstruction described in this invention.

[0052] Figure 3 This is a schematic diagram of the back structure of the smart clothing for fitness motion capture and body posture reconstruction described in this invention.

[0053] Figure 4 This is a schematic diagram illustrating the connection function of various components in the smart clothing for fitness motion capture and body posture reconstruction described in this invention.

[0054] In the diagram: 1: Flexible conductive circuit; 2: Encapsulation layer; 3: Dielectric layer; 4: Working electrode layer; 5: Tight-fitting sportswear top; 6: Tight-fitting sportswear pants; 7: Triboelectric strain sensing fiber for left shoulder joint; 8: Triboelectric strain sensing fiber for right shoulder joint; 9: Triboelectric strain sensing fiber for left elbow joint; 10: Triboelectric strain sensing fiber for right elbow joint; 11: Triboelectric strain sensing fiber for left knee joint; 12: Triboelectric strain sensing fiber for right knee joint; 13: Triboelectric strain sensing fiber for left scapular joint; 14: Triboelectric strain sensing fiber for right scapular joint; 15: Triboelectric strain sensing fiber for back; 16: Encapsulation shell; 17: Data acquisition chip; 18: Power module. Detailed Implementation

[0055] The following description, in conjunction with embodiments and accompanying drawings, illustrates the intelligent clothing for fitness motion capture and body posture reconstruction according to the present invention.

[0056] Example 1

[0057] like Figure 1 As shown, a triboelectric strain sensing fiber is composed of a flexible conductive line 1, an encapsulation layer 2, a dielectric layer 3, and a working electrode layer 4. The dielectric layer is an elastic conduit that completely encapsulates the working electrode layer. The working electrode layer is made of liquid metal. The flexible conductive line is located at both ends of the dielectric layer, and the encapsulation layer is located at the junction of the flexible conductive line and the dielectric layer.

[0058] The encapsulation layer material of the aforementioned triboelectric strain sensing fiber is polydimethylsiloxane (or Ecoflex or UV-curable adhesive), and the liquid metal is a liquid sodium-potassium alloy (or metallic mercury, metallic francium, metallic cesium, metallic gallium, gallium-indium alloy, gallium-tin alloy, gallium-zinc alloy, gallium-indium-tin alloy, or gallium-indium-tin-zinc alloy); the elastic conduit is a silicone tube (or a polyurethane tube or a polyethylene tube). The preparation method of the triboelectric strain sensing fiber is as follows: liquid metal is injected into the elastic conduit through an injection pump until the cavity is filled; flexible conductive lines are inserted from both ends of the elastic conduit; and then the ports of the elastic conduit are sealed with encapsulation material to prevent liquid metal leakage.

[0059] Example 2

[0060] like Figure 2-4 As shown, a smart garment for fitness motion capture and body posture reconstruction includes a wearable device body, a data collection system, and a data processing system. The wearable device body includes a tight-fitting sportswear top 5, tight-fitting sportswear pants 6, a left shoulder joint triboelectric strain sensing fiber 7, a right shoulder joint triboelectric strain sensing fiber 8, a left elbow joint triboelectric strain sensing fiber 9, a right elbow joint triboelectric strain sensing fiber 10, a left knee joint triboelectric strain sensing fiber 11, a right knee joint triboelectric strain sensing fiber 12, a left scapular joint triboelectric strain sensing fiber 13, a right scapular joint triboelectric strain sensing fiber 14, a back triboelectric strain sensing fiber 15, and a packaging shell 16. Each triboelectric strain sensing fiber is the triboelectric strain sensing fiber described in Example 1. The flexible conductive circuit material is silicone wire (or PVC wire, insulated copper wire, insulated aluminum wire, polyester film wire, Teflon wire, etc.). The data collection system includes a data acquisition chip 17 and a power module 18. The data processing system includes a display terminal that executes a comprehensive algorithm. The display terminal is an electronic display device connected via wireless transmission technology, including a mobile phone, computer, display screen, projector, television, or VR glasses. The comprehensive algorithm includes a deep learning algorithm, a multi-dimensional matching calibration algorithm, a feature extraction algorithm, and a body posture reconstruction algorithm. The deep learning algorithm is developed based on a convolutional neural network model, and its main function is to accurately identify the name of fitness exercises. The multi-dimensional matching calibration algorithm is developed based on a local weighted linear regression model, and its main function is to accurately calculate joint bending angles while ignoring differences in user body shape. The feature extraction algorithm is developed based on a machine learning model, and its main function is to determine the macroscopic center of gravity, position, and orientation of the virtual character model. The body posture reconstruction algorithm is developed based on the C# compiled language through the Unity-3D platform, and its main function is to control the joint angles and macroscopic state changes of the virtual character, compare the calculated joint bending angles with the standard bending angles to determine whether the fitness movements are standard, and display the results on the display terminal.

[0061] The wearable device is worn on the surface of the human body. Each of the triboelectric strain sensing fibers and flexible conductive lines is integrated into the tight-fitting sportswear top 5 and tight-fitting sportswear pants 6 through textile technology (spinning, knitting, weaving, embroidery or sewing) to form a wave pattern. The encapsulation shell 16 is set on the tight-fitting sportswear top. There are a total of 9 triboelectric strain sensing fibers, which can capture the bending movements of the human joints and accurately collect motion data. The motion data is transmitted to the data collection system through the flexible conductive lines.

[0062] The data collection system includes a data acquisition chip (a commercially available RJET-14SP chip) and a power module. The data acquisition chip includes a data processor with a detachable data interface and a wireless signal transmission module. The data processor can effectively receive data measured by triboelectric strain sensing fibers and transmit it to the data processing system via the wireless signal transmission module. The data processor and data interface are detachable, facilitating disassembly, maintenance, and repair. The data processor of the data acquisition chip is connected to the triboelectric strain sensing fibers via flexible conductive lines and is integrated into the surface of the tight-fitting sportswear for data transmission. The interface of the flexible conductive lines is a pluggable interface. The data acquisition chip is used to collect data from the triboelectric strain sensing fibers. The data acquisition chip and power module are connected and housed within the encapsulation shell 16, used to receive joint strain data measured by each triboelectric strain sensing fiber and transmit it to the data processing system via the wireless signal transmission module.

[0063] The display terminal of the data processing system is connected to the data acquisition chip via Bluetooth or wireless network signal. It receives data from the triboelectric strain sensing fiber collected by the data acquisition chip. After data analysis by the comprehensive algorithm in its application (APP software or mini program), it displays the body posture configuration and provides real-time fitness feedback suggestions. The data processing system converts the fitness exercise signal into the synchronous movement of the virtual character model through the comprehensive algorithm and displays it on the display terminal. When a person engages in fitness exercises, a deep learning algorithm first identifies the exercise name based on the output voltage signal of the collected triboelectric strain sensing fiber. Then, a multi-dimensional matching calibration algorithm calculates the joint angle data from the output voltage signal. Next, a feature extraction algorithm determines the macroscopic center of gravity, position, and orientation of the virtual character model based on the voltage signal. Finally, a body posture reconstruction algorithm controls the bending angles and macroscopic state changes of each joint of the virtual character based on the previous calculation results, and adjusts the overall posture of the virtual character model based on the macroscopic information. This achieves high-precision body posture reconstruction of fitness exercises in a three-dimensional visualized simulation environment. The actual joint bending angles are compared with standard data. If the actual joint bending angles are greater or less than the standard data, real-time fitness feedback suggestions are provided on the display terminal. This allows the fitness movements to be standardized by increasing or decreasing the joint bending angles. Finally, the display terminal displays the synchronized movement of the virtual character model's fitness exercises.

[0064] The comprehensive algorithm is implemented as an application within the display terminal. Specifically:

[0065] I. The deep learning algorithm includes the following steps:

[0066] (1) Data preparation: The collected raw data is preprocessed, including outlier removal, missing data imputation, filtering, normalization and data segmentation, and the results are integrated into a reliable dataset.

[0067] (2) Signal similarity analysis: Spearman correlation analysis was used to analyze the signal similarity of the collected body postures, calculate the correlation coefficient of any two body posture signals, and statistically analyze the correlation coefficient distribution matrix.

[0068] (3) Training of the 1D-CNN deep learning algorithm model: 80% of the data samples in the reliable dataset were used for training for a total of 300 training cycles, and 20% were used to test the accuracy of the 1D-CNN deep learning algorithm model. The 1D-CNN model contains 4 kernels, 64 filters and 4 convolutional layers.

[0069] (4) Evaluation of 1D-CNN deep learning algorithm model: The efficiency of 1D-CNN model is evaluated by learning accuracy and learning rate, and the reliability of body pose recognition is evaluated by confusion matrix.

[0070] II. The multi-dimensional matching calibration algorithm includes the following steps:

[0071] (1) Basic data preparation: Collect the joint dimensions and height information of several volunteers, and collect body posture signals according to the preset 24 body postures.

[0072] (2) Database establishment: Use support vector regression model to statistically analyze the distribution relationship between joint size and output signal for a specified joint bending angle, and use ordinary regression analysis method to statistically analyze the distribution relationship between joint bending angle and output voltage for a specified joint name, and store the above statistical results in the database.

[0073] (3) Self-calibration: Volunteers collect body posture signals according to the five preset calibration actions. Then, the Spearman correlation analysis method is used to match the distribution relationship model of the size of each joint and the output signal in the database. Next, the two-dimensional database is used to search for the distribution relationship model of the joint bending angle and the output voltage on both sides of the joint size. Finally, local weighted linear regression analysis is used to achieve accurate correction of the individual regression relationship between the joint angle and the output voltage.

[0074] (4) Joint angle calculation: The real-time joint bending angle is obtained by calculating the output signal of each joint in sequence through the corrected individual regression relationship.

[0075] The feature extraction algorithm described in section III includes the following steps:

[0076] (1) Data preparation: Isolate the voltage signals of the knee and back joints in the body posture signal.

[0077] (2) Feature extraction: The decision tree algorithm is used to analyze the temporal features of the knee and back joints and extract features including waveform, trough time, peak time, trough value, peak value, trough amount, number of peaks, high pressure duration, mean, variance, standard deviation, root mean square and kurtosis.

[0078] (3) Spatial element assignment: The center of gravity, position and orientation of the spatial elements of the virtual model are assigned values ​​based on the feature extraction results and statistical height information. The data for assignment are calculated from height using the behavior analysis statistical formula, calculated from height and step length using the spatial geometric model, and estimated from gait analysis statistical results.

[0079] The body pose reconstruction algorithm described in section IV includes the following steps:

[0080] (1) Virtual character model drawing: Fine-tune the preset virtual character model size, skeleton and joint controller position according to the volunteer's height.

[0081] (2) Control program writing: The real-time joint angle data calculated by the multi-dimensional matching calibration algorithm and the center of gravity, position and orientation offset data calculated by the feature extraction algorithm are sent to the Unity-3D platform via Socket communication, and compiled by the C# compiler into joint rotation commands and bone movement commands based on Euler angles.

[0082] (3) Virtual character body posture reconstruction: The skeleton and joint controller executes control commands to drive the virtual character model to change posture.

[0083] In summary, the aforementioned smart clothing for motion capture and body posture reconstruction enables precise full-space motion capture and posture reconstruction of the human body. The wearable human motion capture system is developed based on tight-fitting sportswear. Using textile technology, it integrates triboelectric strain sensors based on the triboelectric effect into the shoulder, scapular, elbow, back, and knee joints of the sportswear. Flexible wires connect the strain sensors to a data acquisition chip. The chip collects sensor data and sends it to a computer. The captured human motion data is used to reconstruct the body posture of a virtual character model through a self-developed algorithm. Simultaneously, the posture configuration of the body during fitness activities can be displayed in real-time in a 3D visualization simulation environment. The sampled data is unaffected by the external environment or experimental setting. The data collected by this invention can represent human fitness postures. The number of triboelectric strain sensors integrated on the tight-fitting sportswear can be increased or decreased according to different needs, and the fitness activities can be reproduced in real-time in a 3D simulation environment, improving the immersion and experience of the fitness activities.

[0084] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A smart garment for fitness motion capture and body posture reconstruction, characterized in that: It includes the wearable device itself, the data collection system, and the data processing system, among which, The wearable device body includes a tight-fitting sportswear, triboelectric strain sensing fibers, flexible conductive circuits, and an encapsulation shell. The triboelectric strain sensing fibers and flexible conductive circuits are integrated into the tight-fitting sportswear using textile technology, and the encapsulation shell is disposed on the tight-fitting sportswear. The data collection system includes a data acquisition chip and a power module. The data acquisition chip and the power module are connected by circuits and disposed within the encapsulation shell. The data acquisition chip is connected to the triboelectric strain sensing fiber through a flexible conductive circuit. The data processing system includes a display terminal that executes a comprehensive algorithm, which includes a deep learning algorithm, a multi-dimensional matching calibration algorithm, a feature extraction algorithm, and a body posture reconstruction algorithm. The display terminal is connected to a data acquisition chip. The deep learning algorithm identifies the motion name based on the output voltage signal. The multi-dimensional matching calibration algorithm calculates the joint angle data from the output voltage signal. The feature extraction algorithm determines the macroscopic center of gravity, position, and orientation of the virtual character model based on the voltage signal. The body posture reconstruction algorithm controls the bending angles and macroscopic state changes of each joint of the virtual character based on the calculation results and provides real-time fitness feedback suggestions. The multi-dimensional matching calibration algorithm includes the following steps: (1) Basic data preparation: Statistics on the 9 joint dimensions and height of several volunteers, and body posture signals were collected according to the 24 preset body postures; (2) Database establishment: Use support vector regression model to statistically analyze the distribution relationship between joint size and output signal for a specified joint bending angle, and use ordinary regression analysis method to statistically analyze the distribution relationship between joint bending angle and output voltage for a specified joint name, and store the above statistical results in the database; (3) Self-calibration: Volunteers collect body posture signals according to the five preset calibration actions. Then, the distribution relationship model between the size of each joint and the output signal is matched in the database according to the Spearman correlation analysis method. Next, the two-dimensional database is used to search for the distribution relationship model between the joint bending angle and the output voltage on both sides of the joint size. Finally, the individual regression relationship between the joint angle and the output voltage is accurately corrected through local weighted linear regression analysis. (4) Joint angle calculation: The real-time joint bending angle is obtained by calculating the output signal of each joint in sequence through the corrected individual regression relationship.

2. The smart clothing for motion capture and body posture reconstruction according to claim 1, characterized in that: The triboelectric strain sensing fiber consists of a flexible conductive line, an encapsulation layer, a dielectric layer, and a working electrode layer. The dielectric layer completely encapsulates the working electrode layer, the flexible conductive line is located at both ends of the dielectric layer, and the encapsulation layer is located at the junction of the flexible conductive line and the dielectric layer.

3. The smart clothing for motion capture and body posture reconstruction according to claim 2, characterized in that: The encapsulation layer is made of polydimethylsiloxane, Ecoflex, or UV-curable adhesive, the dielectric layer is an elastic conduit, and the working electrode layer is made of liquid metal.

4. The smart clothing for motion capture and body posture reconstruction according to claim 3, characterized in that: The liquid metal is mercury, francium, cesium, gallium, gallium-indium alloy, gallium-tin alloy, gallium-zinc alloy, gallium-indium-tin alloy, gallium-indium-zinc alloy, or liquid sodium-potassium alloy; the elastic conduit is a silicone tube, a polyurethane tube, or a polyethylene tube.

5. The intelligent clothing for motion capture and body posture reconstruction according to claim 2, characterized in that: The method for preparing the triboelectric strain sensing fiber involves injecting liquid metal into an elastic conduit using an injection pump until the cavity is filled, inserting flexible conductive lines from both ends of the elastic conduit, and then sealing the ports of the elastic conduit with encapsulation material to prevent liquid metal leakage.

6. The smart clothing for motion capture and body posture reconstruction according to claim 1, characterized in that: The triboelectric strain sensing fiber and flexible conductive circuit are integrated into the tight-fitting sportswear by means of textiles, knitting, weaving, embroidery or sewing. The flexible conductive circuit is made of silicone wire, PVC wire, insulated copper wire, insulated aluminum wire, polyester film wire, Teflon wire, nano silver wire or fluoroplastic wire.

7. The smart clothing for motion capture and body posture reconstruction according to claim 1, characterized in that: The data acquisition chip includes a data processor with a detachable data interface and a wireless signal transmission module. The interface of the flexible conductive line is a pluggable interface, which is used to connect the triboelectric strain sensing fiber and the data processor of the data acquisition chip. It is integrated into the surface of the tight-fitting sportswear for data transmission.

8. The smart clothing for motion capture and body posture reconstruction according to claim 1, characterized in that: The integrated algorithm is installed in the display terminal via an application program, and the data acquisition chip is connected to the application program in the display terminal via Bluetooth or a wireless network.

9. The smart clothing for motion capture and body posture reconstruction according to claim 1 or 8, characterized in that: The display terminal is an electronic display device connected via wireless transmission technology, including mobile phones, computers, displays, projectors, televisions, or VR glasses.

10. The smart clothing for motion capture and body posture reconstruction according to claim 1, characterized in that: The deep learning algorithm includes the following steps: (1) Data preparation: The collected raw data is preprocessed, including outlier removal, missing data imputation, filtering, normalization and data segmentation, and the results are integrated into a reliable dataset; (2) Signal similarity analysis: Spearman correlation analysis was used to analyze the signal similarity of the collected body postures, calculate the correlation coefficient of any two body posture signals, and statistically analyze the correlation coefficient distribution matrix; (3) Training of 1D-CNN deep learning algorithm model: 80% of the data samples in the reliable dataset were used for training for a total of 300 training cycles, and 20% were used to test the accuracy of the 1D-CNN deep learning algorithm model. The 1D-CNN model contains 4 kernels, 64 filters and 4 convolutional layers. (4) Evaluation of 1D-CNN deep learning algorithm model: The efficiency of 1D-CNN model is evaluated by learning accuracy and learning rate, and the reliability of body pose recognition is evaluated by confusion matrix. The feature extraction algorithm includes the following steps: (1) Data preparation: Isolate the voltage signals of the knee and back joints in the body posture signal; (2) Feature extraction: The decision tree algorithm is used to analyze the temporal features of the knee and back joints and extract features including waveform, trough time, peak time, trough value, peak value, trough amount, number of peaks, high pressure duration, mean, variance, standard deviation, root mean square and kurtosis. (3) Spatial element assignment: The center of gravity, position and orientation of the spatial elements of the virtual model are assigned based on the feature extraction results and the statistical height information; the data for assignment are calculated from height based on the behavior analysis statistical formula, calculated from height and position deviation based on the spatial geometric model, and estimated from gait analysis statistical results. The body pose reconstruction algorithm includes the following steps: (1) Virtual character model drawing: Fine-tune the preset size of the virtual character model, the position of the skeleton and joint controllers according to the height of the volunteers; (2) Control program writing: The real-time joint angle data calculated by the multi-dimensional matching calibration algorithm and the center of gravity, position and orientation offset data calculated by the feature extraction algorithm are sent to the Unity-3D platform via Socket communication, and compiled by the C# compiler into joint rotation commands and bone movement commands based on Euler angles; (3) Virtual character body posture reconstruction: The skeleton and joint controller executes control commands to drive the virtual character model to change posture.

Images