Smart sportswear based on somatosensory interaction and injury early warning method

Abstract

The application provides a smart sportswear based on somatosensory interaction and an injury warning method. The smart sportswear comprises: a garment main body having a first region corresponding to a target anatomical site of a human body and a second region corresponding to a human body touch-sensitive area; a sweat guide layer, a sensing layer and a protective layer stacked from inside to outside; a multi-modal sensing unit located in the first region; the multi-modal sensing unit is used for collecting multi-modal sensing signals; the multi-modal sensing signals comprise body surface deformation signals, spatial pose signals and electromyographic signals; a touch feedback unit is located in the second region; a central processing unit is integrated with a warning judgment module; the central processing unit is configured to receive the multi-modal sensing signals, determine whether there is a posture injury risk through the warning judgment module, generate an injury warning signal when it is determined that there is a posture injury risk, and control the touch feedback unit to execute a corresponding vibration mode based on the injury warning signal. The user can obtain an injury warning during the action execution process.

Description

Technical Field

[0001] This application relates to the field of smart wearable device technology, and in particular to a smart sportswear based on motion-sensing interaction and an injury early warning method. Background Technology

[0002] As a product of the deep integration of information technology and sports and health, smart wearable devices have developed rapidly in recent years in the context of national fitness and scientific training. Among them, smart sportswear, with its structural advantages of conforming closely to the human body and integrating multiple types of sensing units, has become an important carrier for real-time monitoring of exercise physiological status, evaluating movement quality, and preventing sports injuries. With the continuous improvement of users' demands for sports safety and scientific training, smart sportswear is accelerating its evolution from the early stage of "basic data collection" towards the intelligent direction of "high-precision biomechanical sensing".

[0003] Currently, mainstream smart sportswear employs a combination of a low-frequency sampling inertial measurement unit (IMU) and a simple thin-film pressure sensor embedded in the garment's main body. The IMU (integrating an accelerometer and gyroscope) records the approximate movement trajectory and posture angles of the limbs at a sampling frequency of 10-50 Hz, while the pressure sensor reflects the magnitude of local contact force through changes in resistance or capacitance, used to determine the force distribution on the soles of the feet or whether the torso remains upright. This type of system is simple in structure, low in cost, and can provide basic motion parameters during static or low-speed movements (such as walking and jogging).

[0004] However, existing sportswear only records or replays abnormal data when it is collected, and cannot generate timely and reliable injury warnings during the execution of movements. In high-intensity and high-complexity fitness scenarios, users often only realize that they have made a mistake after the injury has occurred, which seriously restricts the practical value of smart sportswear in the field of sports protection. Summary of the Invention

[0005] The purpose of this application is to provide a smart sportswear and injury warning method based on somatosensory interaction, addressing the problem that existing sportswear cannot provide timely and reliable sports injury warnings during the execution of movements.

[0006] To address the aforementioned technical problems, this application provides the following technical solutions: The first aspect of this application provides a smart sportswear based on motion-sensing interaction, comprising: The main body of the garment has a first region corresponding to the target anatomical site of the human body and a second region corresponding to the human body's tactile sensitive area; the main body of the garment includes a sweat-wicking layer, a sensing layer and a protective layer stacked from the inside out; A multimodal sensing unit is disposed in the sensing layer and located in the first region; the multimodal sensing unit is used to acquire multimodal sensing signals; the multimodal sensing signals include body surface deformation signals, spatial pose signals, and electromyographic signals; The haptic feedback unit, located in the second region, is configured to generate a variety of different vibration modes; The central processing unit is signal-connected to both the multimodal sensing unit and the tactile feedback unit; the central processing unit integrates a warning determination module. The central processing unit is configured to: receive the multimodal sensing signal, calculate the biomechanical parameters of the target anatomical site based on the multimodal sensing signal through the early warning determination module, determine the risk of posture injury when the biomechanical parameters deviate from the preset safety range, generate an injury early warning signal, and control the tactile feedback unit to execute the corresponding vibration mode based on the injury early warning signal; the biomechanical parameters are selected from at least one of the following: joint angle, radius of curvature, and inter-joint coordination ratio derived from multiple joint angles.

[0007] In some modified embodiments of the first aspect of this application, the first region includes: The first sub-region corresponds to the midline of the human spine, the peak points of the left and right shoulder joints, and the outer sides of the left and right elbow joints. The second sub-region corresponds to the seventh cervical vertebra at the back of the neck and the distal ends of both arms. The third sub-region corresponds to the pectoralis major, deltoid, and rectus abdominis muscles in the human body; The multimodal sensing unit includes: Multiple flexible strain sensors are distributed in the first sub-region; multiple inertial measurement units are distributed in the second sub-region; and multiple electromyography sensors are distributed in the third sub-region.

[0008] In some modified embodiments of the first aspect of this application, the conductive layer of the flexible strain sensor comprises carbon nanotubes and polydimethylsiloxane, wherein the carbon nanotubes are dispersed within the polydimethylsiloxane and form a percolation network; the conductive layer has a microcrack structure. The inertial measurement unit integrates a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer; the inertial measurement unit is configured to operate at a sampling frequency of 500-1000Hz. The electromyography sensor includes a flexible dry electrode array; one end of the flexible dry electrode array is connected to the sensing layer, and the other end passes through the sweat-wicking layer for contact with human skin; the surface of the electrodes in the flexible dry electrode array has micropores with a preset aperture.

[0009] In some modified embodiments of the first aspect of this application, the second region corresponds to both sides of the human spine, the shoulder region, and the elbow region; The haptic feedback unit includes: Multiple linear resonant actuators are distributed in the second region; The biomechanical parameters include the joint angle, the radius of curvature, and the inter-joint coordination ratio; the central processing unit is configured to: When the inter-joint coordination ratio is greater than a first preset ratio threshold and less than or equal to a second preset ratio threshold, a motion rhythm deviation signal is generated, and the tactile feedback unit is controlled to execute a low-frequency continuous vibration mode based on the motion rhythm deviation signal. When the joint angle is greater than the first preset joint threshold and less than or equal to the second preset joint threshold, or when the radius of curvature is greater than the first preset curvature threshold and less than or equal to the second preset curvature threshold, a posture deviation signal is generated, and the tactile feedback unit is controlled to execute a high-frequency intermittent pulse vibration mode based on the posture deviation signal. When the inter-joint coordination ratio is greater than the second preset ratio threshold, or the joint angle is greater than the second preset joint threshold, or the radius of curvature is less than or equal to the first preset curvature threshold, an emergency braking signal is generated, and the haptic feedback unit is controlled to execute a full-array burst vibration mode based on the emergency braking signal. The emergency braking signal has a higher generation priority than the action rhythm deviation signal and the attitude deviation signal.

[0010] In some modified embodiments of the first aspect of this application, the central processing unit further integrates: The signal preprocessing module is configured to preprocess the multimodal sensing signal, the preprocessing including filtering, time synchronization, zero-point drift elimination and feature extraction of different modal signals; The early warning determination module is connected to the signal preprocessing module and is configured to calculate the biomechanical parameters of the target anatomical site based on the preprocessed signal. When the biomechanical parameters deviate from the preset safety range, it is determined that there is a risk of posture damage and the damage early warning signal is generated. The feedback drive module, connected to the early warning determination module, is configured to receive the damage early warning signal and convert it into a corresponding hardware drive instruction, which is then output to the tactile feedback unit to trigger the corresponding vibration mode.

[0011] The second aspect of this application provides a sports injury early warning method based on somatosensory interaction, applied to the smart sportswear described in the first aspect, comprising: S1 acquires multimodal sensing signals, which include body surface deformation signals, spatial pose signals, and electromyographic signals. S2 calculates the biomechanical parameters of the target anatomical site based on the multimodal sensing signals; When the biomechanical parameters deviate from the preset safety range, a risk of posture injury is determined, and an injury warning signal is generated; wherein, the biomechanical parameters include at least one of joint angle, radius of curvature, and inter-joint coordination ratio derived from multiple joint angles; Based on the damage warning signal, S3 controls the tactile feedback unit located in the tactile sensitive area of ​​the garment body to execute the corresponding vibration mode.

[0012] In some modified embodiments of the second aspect of this application, step S1 is further included before: Acquire the baseline sensing signal when performing standard actions; A baseline motion profile is constructed based on the baseline sensing signal, and the baseline motion profile includes the baseline values ​​of the biomechanical parameters of the target anatomical site. Step S2 includes: S201 extracts real-time biomechanical parameters at the target anatomical site from the multimodal sensing signal; S202 performs a matching analysis between the real-time biomechanical parameters and the corresponding baseline values ​​in the baseline motion profile; If the real-time biomechanical parameters deviate from the baseline value beyond the preset safety range, it is determined that there is a risk of posture damage, and a damage warning signal is generated.

[0013] In some modified embodiments of the second aspect of this application, the step of determining the risk of posture damage and generating a damage warning signal when the biomechanical parameters deviate from a preset safety range includes: The preset safety range is determined based on the biomechanical load assessment mechanism; When the inter-joint coordination ratio is greater than the first preset ratio threshold and less than or equal to the second preset ratio threshold, a motion rhythm deviation signal is generated. When the joint angle is greater than the first preset joint threshold and less than or equal to the second preset joint threshold, or when the radius of curvature is greater than the first preset curvature threshold and less than or equal to the second preset curvature threshold, an attitude deviation signal is generated. An emergency braking signal is generated when the inter-joint coordination ratio is greater than the second preset ratio threshold, or the joint angle is greater than the second preset joint threshold, or the radius of curvature is less than or equal to the first preset curvature threshold. The emergency braking signal has a higher generation priority than the action rhythm deviation signal and the attitude deviation signal.

[0014] In some modified embodiments of the second aspect of this application, step S2 includes: S211 acquires the multimodal sensing signals within multiple consecutive motion execution cycles and performs temporal feature extraction; S212 constructs a time-accumulated change curve of attitude deviation based on the aforementioned time-series characteristics; S213 calculates the slope characteristics of the time-series cumulative change curve; If the current slope characteristic exceeds the preset slope warning threshold, it is determined that there is a risk of exceeding the safety limit in the next motion execution cycle, and the damage warning signal is generated in advance.

[0015] In some modified embodiments of the second aspect of this application, the method further includes a tactile teaching mode: Obtain the biomechanical force exertion sequence corresponding to each of the target anatomical sites in the pre-stored standard action model; According to the force exertion sequence, the tactile feedback units distributed at each of the target anatomical sites are activated sequentially to form a temporal vibration wave packet transmitted along the kinetic chain, so as to simulate the muscle force transmission path.

[0016] Compared to existing technologies, the smart sportswear based on somatosensory interaction provided in the first aspect of this application employs a composite structure in the main body of the garment, consisting of a sweat-wicking layer, a sensing layer, and a protective layer layer stacked sequentially from the inside out. This layered design ensures rapid sweat removal and wearing comfort, while also providing a stable fit and physical protection for multimodal sensing units (such as electromyography sensors). This ensures continuous and stable acquisition of multimodal sensing signals and long-term reliable operation of the system, laying a solid foundation for the accuracy and real-time nature of injury warning.

[0017] The multimodal sensing unit is precisely deployed in the first region corresponding to the target anatomical site of the human body (such as the shoulder joint, knee joint, lumbar spine, etc.). The target anatomical site of the human body is a key area where mechanical load is concentrated, posture instability and injury are common during exercise. The multimodal sensing unit can be placed here to directly and efficiently capture local biomechanical features that are highly related to sports injuries, avoiding information attenuation or misjudgment due to positional deviation.

[0018] Simultaneously, the system integrates three types of multimodal sensing data: surface deformation signals, spatial pose signals, and electromyographic signals, constructing a complementary and synergistic high-dimensional sensing system. Surface deformation signals reflect the degree of soft tissue stretching and joint flexion-extension states; spatial pose signals accurately depict the dynamic trajectory and angular relationships of limbs in three-dimensional space; and electromyographic signals reveal the muscle activation sequence, coordination, and force exertion patterns. Multimodal fusion can enhance system robustness through cross-validation, comprehensively covering the three core dimensions of motion control: structure, posture, and neural pathways. This significantly improves the recognition accuracy and warning reliability for complex, high-risk movements (such as knee valgus during squats and arched back during deadlifts).

[0019] The early warning and judgment module is embedded in the central processing unit and configured to directly assess posture damage risk in real time upon receiving multimodal sensing signals. This achieves in-situ fusion of perception and decision-making, significantly reducing system latency. During user actions, immediate, in-vivo assessments of potential posture anomalies or damage risks can be completed, transforming motion safety intervention from traditional "post-event statistical analysis" to "real-time response during the event," achieving a crucial leap from passive retrospection to proactive protection.

[0020] A haptic feedback unit is integrated into the second area of ​​the garment's main body (corresponding to the body's tactile sensitive areas, such as the upper arm, waist, or outer thigh). This unit can generate various differentiated vibration patterns. When the central processing unit determines that there is a risk of postural injury, it immediately generates a warning signal and drives the haptic feedback unit to execute a specific vibration pattern that matches the type or severity of the risk. Through changes in vibration frequency, intensity, and rhythm, the system can intuitively convey different types of risk information, allowing users to perceive the nature of the risk and quickly adjust their actions within milliseconds.

[0021] In summary, the system constructs a complete "perception-decision-execution" closed loop, from multimodal signal acquisition to real-time risk assessment and precise tactile feedback. Users receive corrective prompts during action execution, effectively preventing the repetition and solidification of incorrect action patterns and preventing the accumulation and aggravation of sports injuries from the source. Attached Figure Description

[0022] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, with the same or corresponding reference numerals denoteing the same or corresponding parts, wherein: Figure 1 A schematic diagram of the structure of the smart sportswear of this application is shown from a first-person perspective. Figure 2 A schematic diagram of the structure of the smart sportswear of this application from a second perspective is shown.

[0023] Explanation of icon numbers: 1. Clothing main body; 2. First area; 21. First sub-area; 22. Second sub-area; 23. Third sub-area; 3. Second area; 4. Central processing unit. Detailed Implementation

[0024] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0025] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.

[0026] Example 1

[0027] like Figure 1 and Figure 2 As shown, Embodiment 1 of this application provides a smart sportswear based on motion-sensing interaction, comprising: The main body of the garment 1 has a first region 2 corresponding to the target anatomical site of the human body, and a second region 3 corresponding to the tactile sensitive area of ​​the human body; the main body of the garment 1 includes a sweat-wicking layer, a sensing layer and a protective layer stacked from the inside to the outside. A multimodal sensing unit is disposed in the sensing layer and located in the first region 2; the multimodal sensing unit is used to collect multimodal sensing signals; the multimodal sensing signals include body surface deformation signals, spatial pose signals, and electromyographic signals; The haptic feedback unit, located in the second region 3, is configured to generate a variety of different vibration modes; The central processing unit 4 is signal-connected to the multimodal sensing unit and the tactile feedback unit respectively; the central processing unit 4 integrates an early warning determination module; The central processing unit 4 is configured to: receive the multimodal sensing signal, calculate the biomechanical parameters of the target anatomical site based on the multimodal sensing signal through the early warning determination module, determine the risk of posture injury when the biomechanical parameters deviate from the preset safety range, generate an injury early warning signal, and control the tactile feedback unit to execute the corresponding vibration mode based on the injury early warning signal; the biomechanical parameters are selected from at least one of the following: joint angle, radius of curvature, and inter-joint coordination ratio derived from multiple joint angles.

[0028] Specifically, the main garment 1 serves as the physical carrier of the smart sportswear, conforming to the human body and supporting various components. The overall structure of the main garment 1 can be configured as an upper garment, lower garment, or a combination of upper and lower garments according to the actual application scenario, to adapt to the needs of different types of sports for monitoring parts and feedback areas.

[0029] The main body of the garment 1 adopts a composite functional fabric structure. Among them, the sweat-wicking layer is in direct contact with the human skin, and it can be made of a flexible fabric material with high moisture-wicking properties, such as a knitted fabric made of natural cotton fiber and regenerated cellulose fiber, to quickly wick away sweat generated during exercise, keep the skin dry and improve wearing comfort.

[0030] Preferably, the sweat-wicking layer is woven from modified polyester fibers and spandex fibers, with the outer side of the sweat-wicking layer having a lower porosity than the inner side, forming a differential capillary structure. This differential capillary structure utilizes a capillary force gradient increasing from the inside out to directionally transfer sweat from the skin surface to the side of the sweat-wicking layer away from the skin. Specifically, the sweat-wicking layer is woven from modified polyester fibers and spandex fibers in a specific ratio. The differential capillary structure can be achieved by controlling the fiber density and pore size distribution through weaving processes, making the porosity of the inner side (closer to the skin) of the sweat-wicking layer higher than the outer side (towards the sensing layer), thereby constructing a capillary force gradient increasing from the inside out. Under the action of this differential capillary structure, sweat is rapidly absorbed on the skin-contacting side and directionally migrates along the direction of increased capillary force to the side of the sweat-wicking layer away from the skin, thus preventing sweat accumulation on the skin surface.

[0031] The sensing layer, located between the sweat-wicking layer and the protective layer, serves as the mounting substrate for the multimodal sensing unit. This sensing layer can be made of a flexible electronic composite material based on a thermoplastic polyurethane film. The film thickness can range from 20 to 100 μm, exhibiting high elasticity (elongation at break >300%), good dielectric properties, and environmental stability. On the surface of this substrate, liquid metal (such as gallium indium tin alloy) or conductive silver paste can be used to form a continuous, highly ductile conductive network through laser direct writing, micro-inkjet printing, or screen printing processes. This network is used to carry the acquisition and transmission of body surface deformation, spatial pose, and electromyographic signals.

[0032] Preferably, the sensing layer includes a flexible fabric substrate and a conductive yarn network embedded in the flexible fabric substrate; the conductive yarn network is composed of silver-plated yarns, and the central processing unit 4 is electrically connected to the multimodal sensing unit through the conductive yarn network. Specifically, the conductive yarn network is composed of silver-coated fibers (i.e., silver-plated yarns), which have both high conductivity and excellent tensile durability. This conductive yarn network serves as a pathway for power transmission and signal communication, electrically connecting the multimodal sensing units distributed in the first region 2 of the garment body 1 to the central processing unit 4, thereby achieving low-loss, high-fidelity transmission of multi-source sensing signals.

[0033] The protective layer, located on the outermost layer, can be made of abrasion-resistant, tear-resistant, and somewhat elastic protective materials, such as a polyester microporous membrane composite material with a fluorocarbon coating. This protective layer effectively blocks external moisture, dust, and mechanical friction from interfering with the internal sensing elements, maintaining the overall mechanical strength and aesthetic durability of the garment while retaining necessary breathability to prevent internal heat buildup, thus ensuring long-term stable operation of the system.

[0034] Preferably, the protective layer includes an insulating film that covers and adheres to the outside of the sensing layer to isolate liquid media. Specifically, the protective layer includes an insulating film that is covered and adhered to the outside of the sensing layer by a hot-pressing process. This film can be made of a highly breathable thermoplastic polyurethane material, which can isolate external liquid media (such as rainwater or wash water) from the intrusion of internal electronic components, while maintaining the overall softness and water vapor permeability of the fabric. This structure not only gives the smart sportswear water resistance but also ensures the long-term electrical reliability of the sensing layer in complex usage environments.

[0035] The three-layer structure of the garment body 1 can be integrated in one piece by means of hot pressing, ultrasonic stitching or flexible adhesive, so as to ensure that the electrical connection reliability and structural integrity between the functional layers can be maintained under complex deformation conditions such as stretching and bending, thereby providing a stable and reliable physical support platform for the multimodal sensing unit and tactile feedback unit.

[0036] A multimodal sensing unit is a collection of functional modules integrated into the sensing layer that possess the ability to sense multiple physical or physiological signals. The multimodal sensing unit is composed of a combination of various miniature sensor elements, capable of simultaneously responding to mechanical deformation, spatial orientation changes, and electrophysiological activities. It converts various raw signals into electrical signals that can be analyzed by the central processing unit 4, namely, surface deformation signals, spatial pose signals, and electromyographic signals.

[0037] The multimodal sensing unit can simultaneously acquire signals from multiple sources and transmit the acquired data to the central processing unit 4 in real time via a conductive yarn network integrated within the sensing layer, providing high-dimensional, time-aligned input features for the dynamic assessment of posture damage risk. Surface deformation signals can be acquired via a flexible strain sensor, spatial pose signals via an inertial measurement unit, and electromyography (EMG) signals via an EMG sensor. Multiple of each type of sensor can be used.

[0038] The flexible strain sensor can be embedded inside the sensing layer, with electrode leads at both ends. The electrode leads are reliably electrically connected to the pre-placed conductive yarn network in the sensing layer through conductive adhesive or hot pressing process.

[0039] The inertial measurement unit is detachably mounted on a rigid base on the surface of the sensing layer. The base can be covered with a flexible fabric substrate and has conductive contacts inside. These contacts are connected to the metal pins at the bottom of the inertial measurement unit via a magnetic or snap-fit ​​structure. The wires leading out from the base are electrically connected to a network of conductive yarns.

[0040] The electromyography (EMG) sensor includes a flexible dry electrode array. One end of the flexible dry electrode array can be connected to a conductive yarn network in the sensing layer via conductive silver paste or conductive hot melt adhesive. Micropores or slits can be made in the sweat-wicking layer at the location of the EMG sensor. The electrode contacts of the flexible dry electrode array pass through the micropores of the sweat-wicking layer from one side of the sensing layer and protrude to the inner surface of the sweat-wicking layer (adhering to the surface of human skin). This ensures the mechanical stability of the electrode-skin interface. Furthermore, the sweat-wicking layer effectively maintains the relative dryness of the interface between the skin and the EMG sensor, significantly suppressing contact resistance fluctuations caused by sweat accumulation. This ensures the baseline stability of bioelectrical acquisition such as EMG signals and maintains the reliability of electrical contact, thereby achieving high-quality EMG signal acquisition in high-dynamic motion scenarios.

[0041] All junctions between sensors and the sweat-wicking and sensing layers can be sealed with flexible insulating adhesive to prevent sweat from seeping into the conductive circuitry.

[0042] Target anatomical sites refer to key skeletal or muscular attachment areas in the human body that are indicative of postural control, load distribution, or injury risk during specific movement patterns. The selection of these sites is based on principles of exercise biomechanics to ensure that the acquired signals effectively reflect movement quality and potential risks. Target anatomical sites may include the acromion, greater trochanter of the femur, etc.

[0043] The haptic feedback unit provides real-time, non-invasive tactile feedback to the wearer. Electrically connected to the central processing unit 4 via a network of conductive lines within the sensing layer, the haptic feedback unit is configured to apply mechanical vibration feedback of different modes to the wearer based on damage warning signals output by the warning determination module in the central processing unit 4. The haptic feedback unit can be sandwiched between the sensing layer and the protective layer, or integrated into the inner surface of the protective layer. The haptic feedback unit can be staggered from the multimodal sensing unit to avoid vibration interference with signal acquisition.

[0044] Different vibration modes can be distinguished by at least one parameter among vibration frequency, intensity, duration, and pulse sequence, thereby delivering differentiated and identifiable tactile feedback to the user.

[0045] In some embodiments, the vibration pattern is used to indicate the body part where the injury occurred. For example: A single short tremor (lasting 50ms, frequency 100Hz) indicates "abnormal knee joint movement"; Continuous low-frequency vibration (lasting 2 seconds, frequency 60Hz) indicates "abnormal lumbar flexion"; Intermittent triple tremors (three 50ms pulses, 200ms apart) indicate "abnormal shoulder movement".

[0046] In some embodiments, the vibration mode can be used to reflect the severity of the damage risk, for example: A single short tremor (lasting 50ms, frequency 100Hz) indicates a mild risk; Continuous low-frequency vibration (lasting 2 seconds, frequency 60Hz) indicates moderate risk; Intermittent triple oscillations (three 50ms pulses, 200ms apart) indicate severe risk.

[0047] The system can select one of the strategies to provide tactile feedback based on the application scenario, user settings, or training objectives, ensuring that users can accurately understand the meaning of the warning information.

[0048] Tactile sensitive areas refer to regions on the human body surface with a high density of mechanoreceptors and a low threshold for perceiving vibration or pressure stimuli, such as the lumbosacral region, the deltoid region of the upper arm, or the sternal region. In smart sportswear, tactile feedback units are placed in these tactile sensitive areas to ensure that users can clearly perceive weak vibration signals during exercise, thereby improving the reliability and timeliness of human-computer interaction.

[0049] The central processing unit 4, as the control core of the entire smart sportswear, is configured to receive multimodal sensing signals from the multimodal sensing unit, perform synchronous sampling, signal processing and feature extraction, and determine whether there is a risk of posture damage based on the early warning judgment module; when a risk is determined, a corresponding damage early warning signal is generated, and the tactile feedback unit is controlled to execute the corresponding vibration mode accordingly.

[0050] The central processing unit 4 uses a low-power microcontroller or system-on-a-chip as its hardware core, integrates a high-resolution analog-to-digital conversion interface, and supports high-precision synchronous acquisition of dozens of analog signals from strain sensors, inertial measurement units and electromyography modules through a shared sampling clock.

[0051] When the garment body 1 is a top, a detachable modular housing is located in the upper-middle area of ​​the back of the top, within which the central processing unit 4 is encapsulated. The housing establishes a reliable electrical connection with the circuitry in the sensing layer via magnetic metal contacts. These magnetic metal contacts include a permanent magnet located at the bottom of the housing and a corresponding conductive metal pad, providing both mechanical attraction for repeated assembly and disassembly, and ensuring low-resistance electrical signal transmission. This structure allows users to easily remove the central processing unit 4 entirely before washing, effectively preventing damage to electronic components from washing and extending the system's lifespan.

[0052] The central processing unit 4 synchronously samples the raw signals with the multimodal sensing unit through a high-precision analog-to-digital conversion interface, and drives the vibration actuator of the haptic feedback unit through a digital control interface. It has a built-in low-power wireless communication module (such as Bluetooth LE), supports multi-device networking protocols, and can achieve data alignment and status coordination with supporting protective gear such as smart knee braces and smart sports shoes through a low-latency wireless network, constructing a full-body motion capture network. Simultaneously, the system reserves standardized API interfaces to support the output of anonymized motion feature data to third-party health applications or VR / AR training devices, possessing good openness and ecosystem compatibility.

[0053] The central processing unit 4 can be equipped with a built-in temperature compensation circuit to monitor ambient temperature and human skin temperature in real time, and dynamically adjust the gain parameters of the flexible strain sensor to eliminate signal drift caused by temperature changes and ensure the stability and accuracy of the sensed data.

[0054] The early warning determination module is built into the central processing unit 4 and is configured to perform multi-source data fusion on the surface deformation signal, spatial pose signal and electromyographic signal from the multimodal sensing unit, calculate the biomechanical parameters of the target anatomical site based on the multimodal sensing signal, and determine that there is a risk of posture injury when the biomechanical parameters deviate from the preset safety range, and generate a damage early warning signal corresponding to different posture injury risk situations.

[0055] Biomechanical parameters may include one or more of the following: joint angles, radius of curvature, and inter-joint coordination ratios. Among them: Joint angle refers to the angle formed by two adjacent limb segments around a specific joint rotation axis, such as the knee flexion angle during squatting and the shoulder abduction angle during pressing. It can be obtained by calculating the relative rotation angle after the spatial pose signal collected by the inertial measurement unit is processed by attitude calculation, and combined with the resistance change rate output by the flexible strain sensor located above the corresponding joint, the joint angle is corrected or redundantly estimated through a pre-calibrated mapping relationship to improve the measurement robustness. The radius of curvature is a quantitative geometric parameter that characterizes the degree of local curvature of the spine or trunk contour within a specific motion plane (such as the sagittal plane). It is achieved by fusing the local deformation information of the body surface reflected by the resistance change rate output by flexible strain sensors with the spatial pose signal provided by the inertial measurement unit, and dynamically fitting the spatial coordinates of multiple continuously distributed target anatomical sites. This allows for the real-time reconstruction of the curvature trajectory of the spinal segment and the calculation of its local radius of curvature. This indicator can be used to sensitively identify high-risk postures such as excessive lumbar flexion and thoracic lordosis reversal, providing an objective basis for injury early warning.

[0056] The inter-joint coordination ratio refers to the ratio of the angles of two or more related joints during the movement cycle. For example, the ratio of the hip extension angle to the knee extension angle during a deadlift is used to determine whether the lower limb force chain is coordinated; or the ratio of the humeral abduction angle to the scapular superior rotation angle during shoulder abduction or flexion (i.e., the scapulohumeral rhythm coordination ratio) is used to assess whether the shoulder force chain is coordinated.

[0057] Biomechanical parameters also include composite indicators derived from the fusion of electromyographic signals and flexible strain sensor data, which reflect changes in muscle function (such as fatigue level) and motor control strategies (such as compensatory modes).

[0058] Each biomechanical parameter has a corresponding preset safety range. This preset safety range can be set based on the biomechanical parameter value corresponding to the standard movement performed during the initial calibration phase, or based on a general biomechanical safety threshold determined by a biomechanical load assessment mechanism, or a combination of both. In some embodiments, the preset safety range can also be dynamically corrected by combining the target muscle group activation level or fatigue characteristics collected by the electromyography sensor, or used to identify high-risk states where abnormal muscle compensation exists even though the muscle is within the geometrically safe range.

[0059] In terms of material selection and manufacturing processes, stringent industrial standards are adopted to ensure the long-term reliability of the system under high-intensity sports scenarios. The sensor electrodes have undergone salt spray corrosion testing and more than 5,000 cycles of tensile testing to ensure stable electrical performance under highly acidic sweat environments and repeated deformation conditions. The circuit interconnects use a nano-silver paste dispensing process and are further encapsulated with high-performance elastomer resin, significantly improving impact resistance and fatigue resistance. The clothing is designed based on an ergonomic segmented structure, embedding highly breathable pore fabric in high-heat flow areas such as the armpits and back, effectively enhancing heat dissipation and moisture wicking capabilities, and preventing users from experiencing stuffiness or a feeling of restriction during prolonged wear.

[0060] Meanwhile, the products are manufactured through a highly automated production line: the placement and bonding of sensors are completed by a high-precision robotic arm, ensuring the consistency of position of each sensing point across different batches of products; the sensing layer adopts a vacuum thermoforming process to effectively eliminate interlayer bubbles and improve structural stability under extreme temperature and humidity conditions. Each finished product is verified by a dynamic simulation test bench, simulating multimodal signal responses under extreme movements such as squats, jumps, and twists, ensuring that the products leaving the factory possess industrial-grade reliability and motion adaptability.

[0061] Compared to existing technologies, the smart sportswear based on somatosensory interaction provided in the first aspect of this application adopts a composite structure in the main body of the garment 1, which consists of a sweat-wicking layer, a sensing layer, and a protective layer stacked sequentially from the inside out. This layered design not only ensures rapid sweat wicking and wearing comfort, but also provides a stable fit environment and physical protection for multimodal sensing units (such as electromyography sensors), thereby ensuring the continuous and stable acquisition of multimodal sensing signals and the long-term reliable operation of the system, laying a solid foundation for the accuracy and real-time performance of injury warning.

[0062] The multimodal sensing unit is precisely deployed in the first region 2, which corresponds to the target anatomical site of the human body (such as the shoulder joint, knee joint, lumbar spine, etc.). The target anatomical site of the human body is a key area where mechanical load is concentrated, posture instability and injury are common during exercise. The multimodal sensing unit can be placed here to directly and efficiently capture local biomechanical features that are highly related to sports injuries, avoiding information attenuation or misjudgment due to positional deviation.

[0063] Simultaneously, the system integrates three types of multimodal sensing data: surface deformation signals, spatial pose signals, and electromyographic signals, constructing a complementary and synergistic high-dimensional sensing system. Surface deformation signals reflect the degree of soft tissue stretching and joint flexion-extension states; spatial pose signals accurately depict the dynamic trajectory and angular relationships of limbs in three-dimensional space; and electromyographic signals reveal the muscle activation sequence, coordination, and force exertion patterns. Multimodal fusion can enhance system robustness through cross-validation, comprehensively covering the three core dimensions of motion control: structure, posture, and neural pathways. This significantly improves the recognition accuracy and warning reliability for complex, high-risk movements (such as knee valgus during squats and arched back during deadlifts).

[0064] The early warning and judgment module is embedded in the central processing unit 4 and configured to directly make real-time judgments on posture damage risks upon receiving multimodal sensing signals. This achieves in-situ fusion of perception and decision-making, significantly reducing system latency. During the user's actions, an immediate, in-vivo assessment of potential posture anomalies or damage risks can be completed, transforming motion safety intervention from traditional "post-event statistical analysis" to "real-time response during the event," achieving a crucial leap from passive retrospection to proactive protection.

[0065] A tactile feedback unit is integrated in the second area 3 of the garment body 1 (corresponding to the human body's tactile sensitive areas, such as the upper arm, waist, or outer thigh). This unit can generate various differentiated vibration patterns. When the central processing unit 4 determines that there is a risk of postural injury, it will immediately generate a damage warning signal and drive the tactile feedback unit to execute a specific vibration pattern that matches the type or severity of the risk. Through changes in vibration frequency, intensity, rhythm, etc., the system can intuitively convey different types of risk information, allowing users to perceive the nature of the risk and quickly adjust their actions within milliseconds.

[0066] In summary, the system constructs a complete "perception-decision-execution" closed loop, from multimodal signal acquisition to real-time risk assessment and precise tactile feedback. Users receive corrective prompts during action execution, effectively preventing the repetition and solidification of incorrect action patterns and preventing the accumulation and aggravation of sports injuries from the source.

[0067] When the smart sportswear is worn as a top, such as Figure 1 and Figure 2 As shown, in some embodiments, the first region 2 includes: The first sub-region 21 corresponds to the midline of the human spine, the peak points of the left and right shoulder joints, and the outer sides of the left and right elbow joints. The second sub-region 22 corresponds to the seventh cervical vertebra at the back of the neck and the distal ends of both arms. The third sub-region 23 corresponds to the pectoralis major, deltoid, and rectus abdominis muscles in the human body; The multimodal sensing unit includes: Multiple flexible strain sensors are distributed in the first sub-region 21; the conductive layer of the flexible strain sensor includes carbon nanotubes and polydimethylsiloxane, wherein the carbon nanotubes are dispersed in the polydimethylsiloxane and form a percolation network; the conductive layer has a microcrack structure. Multiple inertial measurement units are distributed in the second sub-region 22; each inertial measurement unit integrates a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer; the inertial measurement unit is configured to operate at a sampling frequency of 500-1000Hz. Multiple electromyography (EMG) sensors are distributed in the third sub-region 23; the EMG sensors include a flexible dry electrode array; one end of the flexible dry electrode array is connected to the sensing layer, and the other end passes through the sweat-wicking layer for contact with human skin; the surface of the electrodes in the flexible dry electrode array has micropores with a preset aperture.

[0068] Specifically, the conductive layer of the flexible strain sensor is composed of carbon nanotubes and polydimethylsiloxane. The carbon nanotubes are uniformly dispersed within the polydimethylsiloxane substrate using ultrasonic dispersion technology. During the molding process, an electric field in a specific direction is applied to induce the formation of an anisotropic percolation conductive network. The conductive layer of the flexible strain sensor can be processed with a microcrack structure through a pre-stretch-release process or a solvent evaporation-induced shrinkage process. This microcrack structure mimics the sensing mechanism of spider suture organs, inducing local disconnection and reconstruction of the conductive pathway even under extremely small strains (e.g., 0.1%-1%), thereby generating a significant resistance jump and greatly improving the sensitivity to capturing minute joint rotation angles.

[0069] This flexible strain sensor utilizes the piezoresistive effect to convert mechanical deformation caused by joint bending or skin stretching into a continuous resistance change signal. After calibration, it maintains a highly linear response within the tensile strain range of 0% to 50%, with hysteresis controlled to within 3%. The central processing unit 4 can poll each strain sensing channel through a multiplexer and quantize the analog signal using a high-resolution analog-to-digital converter to ensure the accuracy and dynamic range of the original signal.

[0070] By deploying these flexible strain sensors in the first sub-region 21—corresponding to the midline of the human spine, the peak points of the left and right shoulder joints, and the lateral positions of the left and right elbow joints—the system can accurately monitor spinal symmetry, shoulder joint range of motion, and elbow flexion and extension, capturing subtle postural deviations in real time during typical fitness movements such as squats, presses, and rows. Utilizing the synergistic effect of the microcrack structure and the anisotropic percolation conductive network, the sensor exhibits high sensitivity in the main deformation direction, while significantly suppressing responses in the vertical or cross directions, effectively reducing signal crosstalk introduced by multi-degree-of-freedom motion coupling. Therefore, the system can provide high-fidelity, low-noise surface deformation signal input to the posture risk warning and judgment module, significantly improving the accuracy and timeliness of sports injury risk identification.

[0071] "The distal ends of the two arms" refers to the area at the junction of the styloid processes of the ulna and radius, used to characterize the spatial movement of the hand and forearm.

[0072] The inertial measurement unit integrates a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, forming a complete nine-degree-of-freedom inertial sensing system. It is configured to operate at a high sampling frequency of 500 Hz to 1000 Hz, enabling precise capture of high-frequency dynamic changes in limbs within three-dimensional space, including linear acceleration, angular velocity, and absolute spatial orientation. Each inertial measurement unit is embedded in the sensor integration layer through a reinforced structure, ensuring stable mechanical fixation and signal output even under vigorous movement or repeated stretching.

[0073] In terms of deployment strategy, the inertial measurement unit (IMU) located at the seventh cervical vertebra in the back of the neck serves as the global reference coordinate system, while IMUs positioned at the distal ends of the left and right arms serve as local coordinate systems. The system effectively isolates the interference of overall torso translation or rotation on upper limb movement analysis by calculating the rotation matrix of the local coordinate system relative to the global reference coordinate system in real time. For example, when a user performs chest expansion or pressing movements while running, this distributed solution logic can accurately identify the true movement trajectory of the upper limbs relative to the torso, avoiding misinterpreting torso swaying as arm movements. This decoupling capability relies on fine compensation for cross-axis interference within the IMU and real-time state estimation of gyroscope random drift (e.g., using an extended Kalman filter algorithm for attitude fusion). By arranging the inertial measurement unit in the second sub-region 22, not only can a high-precision human dynamic skeleton model be constructed, but it can also support the three-dimensional reconstruction of the coordinated movement of the shoulder, elbow and wrist joints, significantly improving the posture analysis ability and the accuracy of movement standardization assessment for complex fitness movements (such as pull-ups, dumbbell presses, battle rope training, etc.).

[0074] The electrode surfaces in the flexible dry electrode array are nanostructured to form a micropore array with a preset pore size, wherein the average pore size of the micropores is two hundred nanometers (i.e., the preset pore size). This micropore structure can effectively lock in trace amounts of conductive sweat during exercise, significantly reducing the contact resistance of the skin-electrode interface without the need for conductive cream. Even in the early stages of exercise when the body has not yet sweated profusely, it can stably acquire electrophysiological signals with a high signal-to-noise ratio.

[0075] The flexible dry electrode array achieves tight coupling with the skin through the taut pressure of the clothing fabric 1, ensuring reliable electrical contact during dynamic movement. The central processing unit 4 is configured to process the acquired surface electromyographic signals in real time: on one hand, it calculates the root mean square (RMS) value of the signal as a quantitative indicator of the current muscle activation intensity; on the other hand, it monitors the downward trend of the median frequency of the signal to determine the depth of muscle fatigue. By comparing the differences in the RMS values ​​of symmetrical muscle groups on both sides (such as the left and right pectoralis majors or the left and right rectus abdominis), the system can identify uneven force exertion and warn of the resulting risk of abnormal lateral spinal force. Furthermore, the long-term accumulated electromyographic data can also be used to assess the improvement in muscle hypertrophy effects and neural drive efficiency, providing a scientific basis for personalized fitness training.

[0076] By deploying such electromyography (EMG) sensors in the third sub-region 23, the activation state and fatigue process of the core upper limb and trunk muscle groups can be accurately monitored, providing real-time feedback on muscle exertion patterns during typical training movements such as presses, push-ups, and crunches. Thanks to the dual design of a nanoporous structure and fabric pressure coupling, this EMG sensing solution combines high stability, comfort, and maintenance-free characteristics, effectively solving the signal drift and inconvenience problems of traditional wet electrodes in high-intensity exercise scenarios, and significantly improving the reliability and practicality of muscle function assessment.

[0077] When the smart sportswear is worn as a top, such as Figure 1 and Figure 2 As shown, in some embodiments, the second region 3 corresponds to the sides of the human spine, the shoulder region, and the elbow region; The haptic feedback unit includes: Multiple linear resonant actuators are distributed in the second region 3; The biomechanical parameters include the joint angle, the radius of curvature, and the inter-joint coordination ratio; The central processing unit 4 is configured as follows: When the inter-joint coordination ratio is greater than a first preset ratio threshold and less than or equal to a second preset ratio threshold, a motion rhythm deviation signal is generated, and the tactile feedback unit is controlled to execute a low-frequency continuous vibration mode based on the motion rhythm deviation signal. When the joint angle is greater than the first preset joint threshold and less than or equal to the second preset joint threshold, or when the radius of curvature is greater than the first preset curvature threshold and less than or equal to the second preset curvature threshold, a posture deviation signal is generated, and the tactile feedback unit is controlled to execute a high-frequency intermittent pulse vibration mode based on the posture deviation signal. When the inter-joint coordination ratio is greater than the second preset ratio threshold, or the joint angle is greater than the second preset joint threshold, or the radius of curvature is less than or equal to the first preset curvature threshold, an emergency braking signal is generated, and the haptic feedback unit is controlled to execute a full-array burst vibration mode based on the emergency braking signal. The emergency braking signal has a higher generation priority than the action rhythm deviation signal and the attitude deviation signal.

[0078] Specifically, the linear resonant actuator can be a miniature linear resonant actuator with a thickness not exceeding 3 mm, ensuring high-density array arrangement without affecting the flexibility and comfort of the garment. Each linear resonant actuator has independent driving capability to achieve precise tactile coding.

[0079] The second preset joint threshold is greater than the first preset joint threshold, the second preset curvature threshold is greater than the first preset curvature threshold, and the second preset ratio threshold is greater than the first preset ratio threshold.

[0080] The central processing unit 4 is configured to dynamically control the driving mode of each micro linear resonant actuator based on the type of damage warning signal generated, realizing a three-level tactile feedback mechanism that matches the risk level and the nature of the deviation. Specifically, the low-frequency continuous vibration mode (e.g., steady-state vibration at 60-80 Hz for 1-3 seconds) simulates a gentle tapping sensation, gently guiding the user to adjust the neuromuscular activation sequence or movement rhythm, avoiding chronic rotator cuff injuries caused by coordination abnormalities; the high-frequency intermittent pulse vibration mode (e.g., 200 Hz, 30% duty cycle, 2-3 pulses per second) provides relatively sharp, discontinuous tactile stimulation with high attention capture, aiming to prompt the user to immediately correct joint alignment or trunk posture, preventing structural damage caused by the accumulation of postural errors; and the full-array burst vibration mode (e.g., 250 Hz, maximum amplitude, 0.5 seconds) has a strong arousal effect with high-intensity, full-range vibration, used to forcibly remind the user of current dangerous movements, effectively preventing acute sports injuries such as ligament tears and herniated discs.

[0081] For example, the physiologically normal range for the scapular rhythm coordination ratio (i.e., the ratio of the humeral abduction angle to the scapular superior rotation angle) during shoulder abduction is approximately 2:1 (i.e., a ratio of 2). A first preset ratio threshold of 1.5 and a second preset ratio threshold of 2.5 can be set. When the real-time measured scapular rhythm coordination ratio is greater than 1.5 and less than or equal to 2.5, a movement rhythm deviation signal is generated, and the tactile feedback unit executes a low-frequency continuous vibration mode to gently remind the user to adjust the shoulder force sequence. When the measured ratio is greater than 2.5, an emergency braking signal is generated, and the tactile feedback unit executes a full-array burst vibration mode to forcibly interrupt the dangerous movement and prevent rotator cuff injury.

[0082] For elbow flexion, the physiologically safe range is typically 0°-150°, with the first preset joint threshold set to 140° and the second preset joint threshold set to 150°. When the elbow flexion angle is greater than 140° but less than or equal to 150°, a posture deviation signal is generated, and the tactile feedback unit executes a high-frequency intermittent pulse vibration mode to prompt the user to control the flexion amplitude. When the flexion angle is greater than 150°, an emergency braking signal is generated, and full-array burst vibration is executed to prevent ligament damage caused by joint hyperextension.

[0083] Regarding the lumbar spine curvature radius, taking the squat as an example, the physiologically safe curvature radius is usually not less than 6 cm. The first preset curvature threshold can be set to 6 cm, and the second preset curvature threshold can be set to 8 cm. When the real-time lumbar spine curvature radius is greater than 6 cm and less than or equal to 8 cm, a posture deviation signal is generated, and high-frequency intermittent pulse vibration is executed to prompt the user to tighten their core and reduce lumbar lordosis. When the curvature radius is less than or equal to 6 cm, an emergency braking signal is generated, and full-array burst vibration is executed to prevent lumbar intervertebral disc overload.

[0084] In some embodiments, each target anatomical site (such as the shoulder, elbow, or lumbar spine region) can be independently equipped with a corresponding linear resonant actuator. When a warning is triggered at a certain anatomical site, only the actuator at that site or its adjacent area executes the corresponding vibration mode, thereby enabling the user to intuitively perceive the specific body part at risk of postural injury and achieve precise spatial tactile feedback guidance.

[0085] The three vibration modes described above transform the biomechanical risk assessment results output by the multimodal sensing system into intuitive tactile codes through a logical mapping mechanism. This coding strategy has been validated by ergonomics, ensuring that users can instantly recognize the feedback semantics and make the correct behavioral response in high-intensity, high-focus fitness scenarios. Simultaneously, the spatial layout of the linear actuators corresponds to the anatomical risk points, achieving both precise local prompts and providing strong global alerts in emergency situations, significantly improving the active protection capabilities and interactive effectiveness of the smart sportswear in mass fitness scenarios.

[0086] In some embodiments, when the smart sportswear is bottoms (e.g., smart sports pants, yoga pants, etc.), the first region 2 and the second region 3 can be set according to the biomechanical characteristics of the lower limbs and the needs of movement function, for example: The first region 2 can correspond to the gluteus maximus region, the peroneus longus region on the outer side of the lower leg, the inner side of the groin, the middle of the front of the thigh (i.e., the region where the rectus femoris muscle is located), the popliteal fossa on the back of the knee joint, the gluteus maximus region, etc., and is used to collect multimodal sensing signals. The second area 3 can correspond to the popliteal fossa behind the knee joint, the gluteus maximus area, etc., and is used to transmit real-time tactile cues to the user through local vibration.

[0087] In some embodiments, the central processing unit 4 further integrates: The signal preprocessing module is configured to preprocess the multimodal sensing signal, the preprocessing including filtering, time synchronization, zero-point drift elimination and feature extraction of different modal signals; The early warning determination module is connected to the signal preprocessing module and is configured to calculate the biomechanical parameters of the target anatomical site based on the preprocessed signal. When the biomechanical parameters deviate from the preset safety range, it is determined that there is a risk of posture damage and the damage early warning signal is generated. The feedback drive module, connected to the early warning determination module, is configured to receive the damage early warning signal and convert it into a corresponding hardware drive instruction, which is then output to the tactile feedback unit to trigger the corresponding vibration mode.

[0088] Specifically, the signal preprocessing module is configured to preprocess multimodal sensing signals from the inertial measurement unit, flexible strain sensor, and electromyography sensor.

[0089] The signal preprocessing module uses an adaptive filtering algorithm to dynamically analyze the signal spectrum distribution and adjust the filter cutoff frequency in real time, effectively eliminating artifact interference caused by large-amplitude limb swings, fabric wrinkles and friction, or body collisions, while retaining key motion feature information.

[0090] To achieve time consistency of multi-source signals, the signal preprocessing module uses a built-in multi-sensor synchronous sampling controller to perform nanosecond-level time alignment of the raw data streams of each sensing module using high-precision clock pulses, and combines a segmented cyclic buffering mechanism to ensure that inertial signals, strain signals and electromyographic signals are strictly synchronized on the same time scale.

[0091] The signal preprocessing module performs a sliding window weighted average operation to compensate for baseline drift in the output of the sensor under static or low dynamic conditions, eliminating zero-point drift. Based on this, it extracts feature vectors containing maxima, zero-crossing rate, first derivative (characterizing motion rate), and second derivative (characterizing acceleration abrupt change) from the preprocessed digital signal as input for subsequent biomechanical analysis.

[0092] The early warning determination module is configured to: construct a real-time human skeleton model based on the feature vector output by the signal preprocessing module, and analyze the instantaneous spatial posture of each major joint; the inertial measurement data in the feature vector is used to calculate the quaternion posture, the resistance change rate of the flexible strain sensor is used to calculate the surface deformation signal, and the electromyographic signal is used to extract the potential timing features to assist in identifying abnormal muscle activation patterns, and to serve as a verification basis for coordination ratio or posture deviation when determining the risk of injury, or to dynamically adjust the preset safety range; the early warning determination module integrates the above multi-dimensional features to calculate the biomechanical parameters of the target anatomical site, including joint angles, relative displacement sequences between adjacent bone segments, spinal curvature radius, and motion coordination ratio, etc.; and compare the biomechanical parameters with the corresponding preset safety ranges respectively; When any parameter exceeds its safe range, the early warning determination module determines the specific damage category, including the damage location and risk level, based on the parameter deviation and the joint, and generates a corresponding damage early warning signal; the type of the damage early warning signal corresponds one-to-one with the preset vibration mode.

[0093] The feedback drive module is configured to receive damage warning signals generated by the warning determination module and, based on the warning type and level, convert them into corresponding hardware drive instructions and output them to the tactile feedback unit. This module supports independent amplitude-frequency control of each miniature linear resonant actuator, enabling precise spatial-temporal tactile encoding.

[0094] The central processing unit 4 may also include a Bluetooth Low Energy (BLE) communication module. The pre-processed digital signal stream is encapsulated and transmitted in real time by the BLE 5.0 or later version. The BLE 4 uses a data packet protocol that includes frame header verification, payload compression, and cyclic redundancy check mechanisms to efficiently and reliably upload the feature-extracted sensor data to external mobile terminals or cloud processing. Even in high electromagnetic interference environments such as gyms, it can ensure extremely low latency and high integrity in data transmission, providing a solid data foundation for subsequent remote 3D motion reconstruction, group movement behavior analysis, and personalized training plan generation.

[0095] In some embodiments, the smart sportswear also includes a hybrid power supply system integrated into the sides and underarm areas of the garment body. The hybrid power supply system includes a flexible triboelectric nanogenerator, a buffer capacitor, a lithium polymer battery, and a smart power management module.

[0096] The flexible triboelectric nanogenerator uses a flexible dielectric material and is placed in a region of intense motion deformation. It utilizes the periodic contact-separation effect caused by human activity to convert mechanical energy into alternating current. This alternating current is converted into direct current by a micro rectifier circuit and then temporarily stored in a buffer capacitor.

[0097] The buffer capacitor is electrically connected between the output of the rectifier circuit and the intelligent power management module to smooth out the pulse fluctuations of the flexible triboelectric nanogenerator output and provide stable auxiliary power supply for the multimodal sensing unit in low-power mode.

[0098] The lithium polymer battery uses flexible packaging and forms a hybrid power supply architecture with a flexible triboelectric nanogenerator, which effectively extends the battery life.

[0099] The intelligent power management system automatically switches working modes according to the motion state: when the inertial measurement unit or flexible strain sensor detects that the user is still for more than a preset time (such as 5 minutes), the system cuts off the power supply to the high-power module and enters nanoampere-level deep sleep; once a specific frequency vibration signal (such as 0.5-5 Hz) that characterizes the start of motion is detected, the system wakes up and resumes full-function operation within milliseconds.

[0100] The top features a flexible wireless charging receiver coil embedded in the area where the brand logo is located on the chest or the decorative label on the shoulder. This placement facilitates alignment with the external charging dock without compromising wearing comfort. The receiver coil is connected to the power management module via a flexible wire, supporting contactless fast charging without the need for a physical interface, effectively improving waterproof performance and long-term reliability.

[0101] Example 2

[0102] Embodiment 2 of this application provides a motion-sensing interaction-based sports injury early warning method, applied to the smart sportswear described in Embodiment 1, with reference to... Figure 1 and Figure 2 ,include: S1 acquires multimodal sensing signals, which include body surface deformation signals, spatial pose signals, and electromyographic signals. S2 calculates the biomechanical parameters of the target anatomical site based on the multimodal sensing signals; When the biomechanical parameters deviate from the preset safety range, a risk of posture injury is determined, and an injury warning signal is generated; wherein, the biomechanical parameters include at least one of joint angle, radius of curvature, and inter-joint coordination ratio derived from multiple joint angles; Based on the damage warning signal, S3 controls the tactile feedback unit located in the tactile sensitive area of ​​the garment body 1 to execute the corresponding vibration mode.

[0103] Specifically, the multimodal sensing signals include surface deformation signals acquired by flexible strain sensors distributed on the main body of the clothing 1, spatial pose signals acquired by inertial measurement units, and electromyographic signals acquired by surface electromyography sensors.

[0104] The multimodal sensing signals are preprocessed, and based on the preprocessed signals and the biomechanical load assessment model, it is determined whether there is a risk of posture damage; if the posture damage risk is determined to exist, a damage warning signal is generated. Based on the damage warning signal, the tactile feedback unit located in the preset second area 3 of the main body of the garment 1 is controlled to execute a vibration mode corresponding to the type or level of posture damage risk.

[0105] In some embodiments, the method further includes the following steps prior to step S1: Acquire the baseline sensing signal when performing standard actions; A baseline motion profile is constructed based on the baseline sensing signal, and the baseline motion profile includes baseline values ​​of the biomechanical parameters of the target anatomical site.

[0106] Specifically, prior to step S1, the system performs an initial calibration process to establish a baseline motion profile. This baseline motion profile serves as a reference for posture risk assessment during subsequent training. Its construction supports two modes, which users can choose according to their needs: Mode 1 is based on a pre-built library of national fitness standard models.

[0107] The system has a pre-built standardized multimodal reference sensing signal library for common fitness movements (such as squats, shoulder abductions, lunges, and presses). This signal library was constructed by professional institutions based on inertial, strain, and electromyographic signals collected from a large number of healthy individuals performing standard movements. Through statistical analysis and kinematic reconstruction, a reusable reference dataset was created. In this mode, users do not need to perform calibration movements on-site. The system retrieves the corresponding reference sensing signal representing the standard movement from the signal library according to the selected training movement type; subsequently, it constructs the user's reference movement profile based on the reference sensing signal.

[0108] Mode 2: Based on user-personalized calibration data.

[0109] Under the guidance of a coach or video, the user completes several sets of typical training movements under standard instruction (such as standard squats, shoulder abduction, and lunges). During this process, multimodal sensing units distributed on the main body 1 of the smart clothing—including inertial measurement units, flexible strain sensors, and electromyography sensors—simultaneously collect surface deformation signals, three-dimensional spatial pose signals, and electromyography signals as reference sensing signals for performing standard movements. Based on the multimodal reference sensing signals, the central processing unit 4 establishes a global coordinate system using the inertial measurement unit at the seventh cervical vertebra as a global reference point. It then uses a rotation matrix to real-time remove interference from the overall trunk movement on the local limb movements, constructing a high-precision dynamic human skeleton model. Based on this model, it generates a user-specific reference movement contour and sets a corresponding personalized preset safety range (e.g., reference value ± dynamic safety interval) based on the biomechanical parameter reference values ​​of the target anatomical sites extracted from the reference movement contour. The preferred embodiment of this application is Mode Two.

[0110] Accordingly, step S2 includes: S201 extracts real-time biomechanical parameters at the target anatomical site from the multimodal sensing signal.

[0111] S202 performs a matching analysis between the real-time biomechanical parameters and the corresponding baseline values ​​in the baseline motion profile; If the real-time biomechanical parameters deviate from the baseline value beyond the preset safety range, it is determined that there is a risk of posture damage, and a damage warning signal is generated.

[0112] Specifically, the matching analysis can be performed in the central processing unit 4, the mobile terminal, or the cloud processing center. It assesses the deviation between real-time biomechanical parameters and a preset safety range based on a baseline value using a pre-defined logical decision matrix or a biomechanical standard model library. If the deviation exceeds the preset safety range, a risk of posture damage is determined.

[0113] In some embodiments, the baseline motion profile further includes dynamic kinematic features. This baseline motion profile is based on multimodal sensor data of standard motion samples. Through an individualized anatomy-sensor mapping model or inverse kinematics algorithm, its underlying feature vectors (including the first derivative reflecting local motion rate, the second derivative characterizing acceleration abrupt changes, and zero-crossing rates and maxima used to identify the onset and cessation of electromyographic activation / inhibition) are fused and reconstructed to generate and solidify high-level baseline kinematic parameters (e.g., joint angle timing, spinal segment curvature, multi-muscle group synergistic activation phase, etc.).

[0114] During real-time motion monitoring, the system performs the same feature extraction and reconstruction process on the currently acquired multimodal sensing signals to generate corresponding real-time high-level kinematic parameters, and then performs matching analysis with the high-level baseline kinematic parameters. This mechanism significantly improves the accuracy, real-time performance, and adaptability to individual differences in posture injury risk identification.

[0115] In some embodiments, matching analysis supports digital grading of athletic performance: the central processing unit 4 automatically calculates the quality score of a single training session based on the degree of conformity between the movement and the baseline contour, the rhythmic smoothness, and the energy efficiency, and uploads the score along with the historical physical fitness evolution curve to the cloud processing center to identify the user's long-term weaknesses in specific muscle group strength or joint flexibility, thereby providing dynamic optimization suggestions for subsequent training.

[0116] In some embodiments, determining the risk of posture damage when the biomechanical parameters deviate from a preset safety range and generating a damage warning signal includes: The preset safety range is determined based on the biomechanical load assessment mechanism; When the inter-joint coordination ratio is greater than a first preset ratio threshold and less than or equal to a second preset ratio threshold, a motion rhythm deviation signal is generated.

[0117] An attitude deviation signal is generated when the joint angle is greater than a first preset joint threshold and less than or equal to a second preset joint threshold, or when the radius of curvature is greater than a first preset curvature threshold and less than or equal to a second preset curvature threshold.

[0118] An emergency braking signal is generated when the inter-joint coordination ratio is greater than the second preset ratio threshold, or the joint angle is greater than the second preset joint threshold, or the radius of curvature is less than or equal to the first preset curvature threshold.

[0119] The emergency braking signal has a higher generation priority than the action rhythm deviation signal and the attitude deviation signal.

[0120] Specifically, the inter-joint coordination ratio can include the scapulohumeral rhythm coordination ratio in shoulder abduction movements, the thoracolumbar coordination curvature ratio during trunk flexion (i.e., the ratio of the curvature changes of the thoracic and lumbar segments), etc.

[0121] Taking the scapular rhythm coordination ratio as an example, the system can estimate the humeral abduction angle and scapular superior rotation angle based on the surface deformation signals collected by flexible strain sensors distributed in the first sub-region 21 at the peak points of the left and right shoulder joints, combined with the three-dimensional spatial pose signals output by the inertial measurement units located at the seventh cervical vertebra in the back of the neck and in the second sub-region 22 at the distal ends of both arms. The system calculates the ratio of the humeral abduction angle to the scapular superior rotation angle in real time as the current scapular rhythm coordination ratio. If this ratio deviates from the preset physiological coordination ratio range and exceeds the individualized dynamic safety range established based on user calibration data, it is determined that there is a risk of rotator cuff muscle group load imbalance, and a corresponding injury warning signal is generated. For example, when the real-time measured scapular rhythm coordination ratio is greater than 1.5 and less than or equal to 2.5, a movement rhythm deviation signal is generated; when the measured ratio is greater than 2.5, an emergency braking signal is generated, and the tactile feedback unit executes a full-array burst vibration mode to forcibly interrupt the dangerous movement and avoid rotator cuff injury.

[0122] Joint angles can include elbow flexion, knee flexion (a key parameter reflecting lower limb load distribution during squats or lunges), and hip abduction (a key indicator affecting pelvic stability during lateral leg raises or single-leg standing).

[0123] Taking the elbow joint as an example, the system calculates the flexion angle of the elbow joint in real time based on the skin stretching deformation signal collected by the flexible strain sensor distributed in the first sub-region on the outer side of the left and right elbow joints, combined with the angular velocity and acceleration data output by the inertial measurement unit in the second sub-region of the ipsilateral upper arm or forearm.

[0124] Specifically, the elbow joint flexion and extension state is inverted by the resistance change rate of a flexible strain sensor, and cross-validation and noise suppression are performed using dynamic kinematic information provided by an inertial measurement unit, thereby obtaining a highly robust joint angle estimate. The system compares the real-time elbow joint flexion angle with preset two-level safety thresholds: if the angle is greater than the first preset joint threshold (e.g., 140°) and less than or equal to the second preset joint threshold (e.g., 150°), a posture deviation signal is generated; if the angle exceeds the second preset joint threshold (e.g., greater than 150°), it is determined that there is a risk of ligament overstretching or bony structure impact, and an emergency braking signal is generated.

[0125] The radius of curvature can include the radius of curvature of the lumbar spine, the radius of curvature of the thoracic spine (such as during bent-over rows or bench presses, used to monitor whether the upper back is excessively rounded), and the radius of curvature of the cervical spine (such as during lat pulldowns or shoulder presses, to assess the risk of head extension and cervical load).

[0126] Taking the lumbar spine curvature radius as an example, the system calculates the lumbar spine segment curvature radius in real time based on the multi-point surface deformation signals collected by the flexible strain sensor array in the first sub-region distributed along the midline of the spine, combined with the overall pose reference provided by the inertial measurement unit in the second sub-region of the trunk.

[0127] Specifically, the high sensitivity of the microcrack structure of the flexible strain sensor is utilized to obtain the continuous strain distribution in the lumbar region; by fitting the strain gradient between adjacent sensing points, the local curvature field is reconstructed, and the equivalent radius of curvature corresponding to the current lumbar lordosis is calculated accordingly.

[0128] The system compares the real-time radius of curvature with preset dual-level safety thresholds: if the radius of curvature is greater than the first preset radius of curvature (e.g., 6 cm) and less than or equal to the second preset radius of curvature (e.g., 8 cm), a posture deviation signal is generated to prompt the user to tighten their core muscles and reduce lumbar lordosis; if the radius of curvature is less than or equal to the first preset radius of curvature (e.g., ≤6 cm), it is determined that the lumbar spine is in a state of excessive lordosis or high shear load, and there is a risk of intervertebral disc mechanical overload, and an emergency braking signal is generated.

[0129] By setting a preset safety range based on the biomechanical load assessment mechanism, the assessment of injury risk can be transformed from subjective experience judgment into quantifiable and repeatable objective threshold control, effectively avoiding acute and chronic sports injuries caused by joints / spines exceeding their physiological tolerance limits.

[0130] Furthermore, this preset safety range can be combined with the biomechanical parameter baseline values ​​of the target anatomical site extracted from the user's baseline movement profile: the joint angles, coordination ratios or radii of curvature collected when an individual completes a typical movement (such as squatting or shoulder abduction) under standard, painless, and controlled conditions are used as a personalized baseline. On this basis, a floating safety tolerance range (e.g., ±10%) is dynamically expanded, thereby achieving a dual-mode safety boundary setting that takes into account both anatomical universality and individual differences---preserving the protective role of general biomechanical guidelines while adapting to the unique performance of different users in terms of flexibility, muscle strength, and neural control.

[0131] The generated damage warning signal is received by the feedback drive module and converted into corresponding hardware drive instructions, controlling multiple linear resonant actuators distributed in the second region 3 to execute differentiated vibration modes: motion rhythm deviation signal → low-frequency continuous vibration; posture deviation signal → high-frequency intermittent pulse vibration; emergency braking signal → full-array burst vibration. This mechanism is formed by the coordinated operation of the signal preprocessing module, warning judgment module and feedback drive module integrated in the central processing unit 4, forming a "perception-assessment-feedback" closed loop, providing real-time, individualized and biomechanically based active protection capabilities for national fitness scenarios.

[0132] In some embodiments, step S2 includes: S211 acquires the multimodal sensing signals within multiple consecutive motion execution cycles and performs temporal feature extraction.

[0133] Specifically, the system, through an inertial measurement unit, flexible strain sensor, and surface electromyography (EMG) sensor integrated into the smart training garment, continuously collects multimodal sensing signals for at least three complete motion cycles during the user's repeated execution of the same training movement (such as squats, bench presses, or lunges). The central processing unit 4 performs time alignment and noise reduction on the raw signals, extracting key temporal features within each cycle, including but not limited to: the time series of joint angles, the first and second derivatives (velocity) and acceleration of the motion trajectory, the rate of change of surface strain, the intra-cycle distribution of the zero-crossing rate of the EMG signal, and the decay rate of the median frequency. These features collectively constitute a dynamic index set reflecting movement stability and control ability.

[0134] S212 constructs a time-accumulated change curve of attitude deviation based on the aforementioned time-series characteristics.

[0135] Specifically, the system compares the kinematic features extracted in each motion cycle with the user's personalized baseline motion profile point by point, calculating the deviation (e.g., angle deviation, activation timing offset, or strain distribution asymmetry) at each key node (such as the maximum knee flexion angle, the peak time of lumbar lordosis, etc.). Then, the deviations corresponding to each cycle are accumulated or weighted averaged in chronological order to generate a "chronological cumulative change curve of posture deviation" reflecting the trend of posture control ability degradation. This curve can visually show whether the motion compensation is showing a continuous deterioration trend.

[0136] S213 calculates the slope characteristics of the time-series cumulative change curve.

[0137] If the current slope characteristic exceeds the preset slope warning threshold, it is determined that there is a risk of exceeding the safety limit in the next motion execution cycle, and the emergency braking signal is generated in advance.

[0138] Specifically, the system performs local linear fitting on the aforementioned cumulative change curve or uses the sliding window difference method to calculate its slope (i.e., the rate of change of the deviation increment within a unit period) in real time. This slope characteristic characterizes the degree of acceleration of attitude instability—the larger the slope, the faster the motion control capability is declining.

[0139] If the currently calculated slope characteristic exceeds the preset slope warning threshold (which can be dynamically adjusted based on training intensity, movement type, and user's historical fatigue data), the system determines that even if the current cycle has not yet exceeded the static safety boundary, it is highly likely to exceed the biomechanical safety limits (such as ligament stretching limits, intervertebral disc shear force thresholds, etc.) in the next exercise execution cycle. At this time, the system immediately generates an emergency braking signal and issues a strong intervention command to the user through the feedback module of the intelligent training device (such as a vibration motor, voice prompts, or automatic locking devices of linked training equipment) to forcibly stop the current training movement and prevent acute injury.

[0140] The slope warning threshold can be: During squatting, the knee joint angle deviation is greater than 3.0° / cycle; the lumbar spine curvature radius reduction is less than -0.8 cm / cycle; During the shoulder press movement, the scapulohumeral synergy ratio deviates by more than 15% per cycle; During the high pulldown exercise, the cervical spine curvature radius decreases by less than -0.5 cm per cycle; During the lunge motion, the hip abduction angle deviation is greater than 2.5° per cycle.

[0141] Positive slope indicates an increase in deviation, while negative slope is used for parameters where "the smaller the slope, the more dangerous it is" (such as the radius of curvature). The threshold can be dynamically adjusted with fixed step sizes (such as 0.5° / cycle, 2% / cycle, 0.1 cm / cycle) based on training intensity, fatigue state, etc.

[0142] For example, during a squat, the system calculates the actual deviation for each cycle in real time based on a personalized baseline established during the user calibration phase (maximum knee flexion angle of 110°): Cycle 1: actual angle 112°, deviation 2.0°; Cycle 2: actual angle 116°, deviation 6.0°; Cycle 3: actual angle 121°, deviation 11.0°. Although the deviation in Cycle 3 (11.0°) is still below the static safety threshold (e.g., 12°), linear fitting of these three cycles yields a slope of 4.5° / cycle, significantly exceeding the preset slope warning threshold (3.0° / cycle). Based on this, the system predicts that the deviation in Cycle 4 will be approximately 15.5°, exceeding the static safety threshold. Therefore, before Cycle 4 begins and Cycle 3 ends, the system immediately generates an emergency braking signal to prevent the next squat, effectively avoiding potential injuries caused by loss of control and truly achieving "prevention is better than cure." By combining multi-cycle time-series modeling with slope dynamic analysis, the risk warning is upgraded from "threshold exceeding the limit alarm" to "trend loss of control prediction". This achieves the technical effect of generating an emergency braking signal before the start of the next action cycle, significantly improving the active protection capability and user safety of the intelligent fitness system.

[0143] In some embodiments, the method further includes a tactile teaching mode: Obtain the biomechanical force application timing sequence corresponding to each target anatomical site in the pre-stored standard action model.

[0144] Specifically, the standard movement model can utilize a national fitness standard model library or user-personally calibrated data. This standard movement model not only includes the spatial location and movement trajectory of each target anatomical site (such as the quadriceps, gluteus maximus, erector spinae, latissimus dorsi, etc.), but also needs to accurately record the activation initiation time, peak time, and duration of each muscle group within the movement cycle, i.e., the "biomechanical force exertion sequence." For example, in a standard squat, the force exertion sequence is typically: Initial phase: The center of pressure on the sole of the foot shifts backward → slight activation of the gastrocnemius muscle; Early squat: The gluteus maximus and hamstrings initiate the movement to stabilize the hip joint; Mid-squat to the lowest point: The quadriceps dominate knee flexion control; Ascending phase: The gluteus maximus and quadriceps work together, while the erector spinae maintains trunk rigidity.

[0145] The timing data of the force exertion is stored in the form of a timestamp sequence in a local or cloud-based action library, corresponding one-to-one with each target anatomical site.

[0146] According to the force exertion sequence, the tactile feedback units distributed at each of the target anatomical sites are activated sequentially to form a temporal vibration wave packet transmitted along the kinetic chain, so as to simulate the muscle force transmission path.

[0147] Specifically, the garment integrates multiple tactile feedback units, which are precisely positioned on the body surface corresponding to the target anatomical locations in the standard motion model, such as: the front of the thigh (rectus femoris), the outer side of the buttock (gluteus medius), the lower back (erector spinae), the upper back (lower latissimus dorsi), and the back of the calf (gastrocnemius). Each tactile unit has independent control capabilities and can adjust the vibration intensity, frequency, and duration.

[0148] After the tactile teaching mode is activated, the system reads the standard force sequence data corresponding to the currently selected training movement and sequentially activates the tactile feedback units at each anatomical location, forming a temporal vibrational packet propagating along the human kinetic chain. This packet simulates the transmission path and rhythm of real muscle activation, for example: When demonstrating the "deadlift" movement: t=0ms: Slight vibration of the foot and posterior lower leg unit (indicating center of gravity distribution and ankle stability); t=150ms: Synchronous vibration of the gluteus maximus and hamstring regional units (indicating the initiation of force through a hip-joint-dominated hinged hip flexion movement). t=300ms: Continuous vibration in the lower back and core area (indicating maintenance of neutral spinal position); t=450ms: Upper back and grip strength related units are activated (indicating scapular stability and tension transmission).

[0149] The time intervals, durations, and intensities of each vibration event strictly match the electromyographic activation sequence in the standard model, forming a perceptible "force-guided flow".

[0150] During static or slow-moving actions, users perceive the transmission sequence of vibrational wave packets through tactile sensation, thereby internalizing the correct force application sequence and muscle coordination patterns. This multi-point temporal tactile stimulation effectively activates the proprioceptive system, promotes the encoding of correct movement procedures in the motor cortex, and is particularly suitable for beginners to establish basic movement patterns, rehabilitation patients to rebuild neuromuscular control, and high-level athletes to optimize force application efficiency.

[0151] The system can also combine real-time motion capture data to compare the synchronization between the user's actual force application timing and the standard vibration wave packet, and provide feedback (such as "the hips started too late") to achieve closed-loop learning.

[0152] This tactile teaching mode transforms the abstract biomechanical force sequence into a tangible, sequential vibration experience, enabling non-visually dependent movement instruction (suitable for training with eyes closed, in dark environments, or in distracted situations); intuitive perception of muscle activation sequence, reducing cognitive load; and proactive construction of kinetic chain coordination, improving movement quality and training safety. It significantly enhances the naturalness of human-computer interaction and the effectiveness of teaching.

[0153] In some embodiments, a remote exercise guidance and health management method based on multi-source sensor data is also included, specifically comprising the following steps: 3D motion reconstruction and real-time visualization: Receives multimodal sensing signals from smart sportswear; generates high-fidelity 3D motion reconstruction animation based on human kinematics model; maps the real-time activation intensity of each muscle group and the force distribution of joints in the animation through color gradient mapping, and displays motion indicators simultaneously, including step frequency per minute, joint range of motion, and power consumption of a single repetitive movement.

[0154] Safety warning and local offline identification: A lightweight judgment model runs in the central processing module of the terminal device or wearable unit; when a high-risk physiological or mechanical event (such as severe arrhythmia, excessive joint shear force, or signs of acute sprain or contusion) is detected, a local alarm mechanism is triggered even without a network connection to ensure the bottom line of sports safety.

[0155] Personalized training plan generation: Upload users' physical condition data, historical exercise records, and fatigue feedback to the cloud; based on a big data analysis engine, combined with their age, gender, fitness level, and other characteristics, automatically optimize training content and intensity, generate and push weekly personalized training plans.

[0156] Group benchmarking and weakness identification: On the cloud platform, the user's performance is compared with that of a group with similar characteristics; weaknesses in specific muscle group strength, joint flexibility or movement coordination are identified; and targeted strengthening exercises are dynamically inserted into the subsequent training plan.

[0157] Preventive health monitoring: For sub-healthy users, it combines heart rate variability and respiratory quotient analysis; it judges whether the current cardiopulmonary load is within a safe range based on preset logic rules; if it exceeds the threshold, it adjusts the training intensity or suggests suspending exercise to achieve proactive health intervention.

[0158] Medical-grade rehabilitation parameter execution: Receives restricted parameters set by remote rehabilitation physicians or professional coaches (such as the weight-bearing capacity of the affected limb after surgery not exceeding 30% of body weight, the upper limit of joint range of motion, etc.); converts these parameters into sensor monitoring rules; compares the actual data in real time during the user's home training process, and triggers graded intervention measures once the limit is exceeded.

[0159] Group knowledge mining and model evolution: Aggregate massive amounts of anonymous user exercise data and classify them according to demographic characteristics; identify common mistakes and high-incidence injury patterns that different groups of people are prone to make during fitness through in-depth mining; feed the obtained patterns back to the local early warning and scoring model to achieve continuous iteration and optimization of the algorithm.

[0160] Digitalization of athletic performance and motivational feedback: Automatically calculates a single training session quality score based on movement accuracy, rhythm consistency, and energy utilization efficiency; uses the score results as a reference for user self-improvement, and supports social comparison within authorized scope to enhance training adherence and motivation.

[0161] In some embodiments, the method further includes a kinematic chain linkage constraint correction step: Based on the human biomechanical kinetic chain model, the upper limbs, trunk, and lumbar spine are considered as a linked whole. When a slight deviation in the posture of a local joint is detected, the system automatically determines whether the deviation is caused by abnormal movement of adjacent upstream or downstream joints. If it is determined to be a kinetic chain transmission deviation, the posture or force application sequence of the upstream joint is dynamically corrected in advance, and targeted force application mode correction prompts are output to the user to prevent the deviation from amplifying step by step along the kinetic chain and to prevent cumulative damage.

[0162] Specifically, this step is based on the closed / open kinetic chain theory in human biomechanics, and constructs a multi-segment linkage model of upper limb-trunk-lumbar spine, treating the shoulder joint, elbow joint, thoracic spine, lumbar spine and other parts of the spine as a dynamic whole working together.

[0163] When analyzing user movements in real time, the system not only assesses whether the angles or torques of individual joints exceed the normal range, but also analyzes the source of local posture deviations through kinetic chain transfer functions. For example, when it detects a slight increase in lumbar lordosis (i.e., lumbar hyperextension) during a user's squat, the system does not only consider it a lumbar problem, but also retrospectively analyzes whether hip joint range of motion, knee joint flexion angle, and ankle joint dorsiflexion ability are limited. If it finds that insufficient hip joint flexion leads to compensatory trunk backward extension, which in turn causes increased lumbar shear force, then the lumbar deviation is determined to be a tracing deviation caused by upstream hip joint dysfunction.

[0164] In this situation, the system will activate the linkage correction mechanism: on the one hand, it will highlight the message "The starting point of force should be initiated from the hip" in the 3D motion reconstruction interface, and indicate the correct hip hinge movement trajectory through arrows or dynamic guide lines; on the other hand, it will automatically insert auxiliary exercises (such as dynamic hip flexor stretching, glute bridge activation, etc.) targeting hip joint flexibility and gluteal muscle activation into the subsequent training plan to improve the coordination of the kinetic chain from the source.

[0165] This linkage constraint correction mechanism effectively avoids the limitations of traditional local interventions that only target the symptom sites. By identifying and correcting the compensatory sources upstream of the kinetic chain, it blocks the amplification of erroneous movement patterns along the biomechanical chain, thereby significantly reducing the risk of chronic strain or cumulative injury caused by long-term postural compensation.

[0166] Example 1

[0167] The smart sportswear (top) of this invention uses a conductive yarn network in its sensing layer made of silver-plated fibers with a diameter of 5 micrometers. These fibers are silver-coated conductive fibers. The sportswear integrates a flexible strain sensor, an inertial measurement unit, and an electromyography sensor, with a central processing unit 4 having a sampling rate of 1000 Hz.

[0168] Comparative Example 1 The device uses commercially available smart fitness apparel, which integrates only a low-frequency inertial measurement unit and a simple pressure sensor (without electromyography and flexible strain sensors), with a sampling frequency of 50 Hz.

[0169] Both groups of programs were tested under the same training conditions: ambient temperature 28℃, relative humidity 65%, users performed high-intensity training including standard squats and various non-standard induced movements (such as knee valgus, excessive forward tilting of the torso, etc.) for 30 minutes, accompanied by profuse sweating. The performance test results of Example 1 and Comparative Example 1 are shown in Table 1.

[0170] Table 1 Performance comparison between Example 1 and Comparative Example 1

[0171] As shown in Table 1, Example 1, through the fusion of a high sampling rate of 1000 Hz and multiple source sensors, can accurately calculate joint angles (error ±0.5°), with motion recognition latency of less than 15 ms, meeting the requirements for real-time biomechanical feedback; Under conditions of high-intensity perspiration, the present invention effectively suppresses signal drift caused by sweat (<2%) through the synergistic design of the sweat-wicking layer, the protective layer and the stable conductive yarn network; while the comparative example, lacking a protective structure, shows serious signal distortion after 20 minutes. This invention is based on the fusion of electromyography and strain data, which can identify the movement compensation caused by muscle fatigue 45 seconds in advance, and guide the user to correct the posture through tactile feedback, with a correction success rate of 92.8%; the comparative example, due to its single function, does not have this capability at all. After 50 standard washes, the resistance change of the sensor of this invention is less than 1.5%, indicating that its fabric electronic structure has good mechanical and chemical stability and is suitable for long-term daily use.

[0172] In summary, this invention not only has significant engineering advantages in terms of measurement accuracy, response speed, and environmental robustness, but also achieves a technological leap from "passive recording" to "active protection" through multimodal perception and intelligent feedback mechanisms. It is especially suitable for fitness and rehabilitation scenarios where there are high requirements for movement safety and training effectiveness.

[0173] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart sportswear based on motion-sensing interaction, characterized in that, include: The main body of the garment has a first region corresponding to the target anatomical site of the human body and a second region corresponding to the human body's tactile sensitive area; the main body of the garment includes a sweat-wicking layer, a sensing layer and a protective layer stacked from the inside out; A multimodal sensing unit is disposed in the sensing layer and located in the first region; the multimodal sensing unit is used to acquire multimodal sensing signals; the multimodal sensing signals include body surface deformation signals, spatial pose signals, and electromyographic signals; The haptic feedback unit, located in the second region, is configured to generate a variety of different vibration modes; The central processing unit is signal-connected to both the multimodal sensing unit and the tactile feedback unit; the central processing unit integrates a warning determination module. The central processing unit is configured to: receive the multimodal sensing signal, calculate the biomechanical parameters of the target anatomical site based on the multimodal sensing signal through the early warning determination module, determine the risk of posture injury when the biomechanical parameters deviate from the preset safety range, generate an injury early warning signal, and control the tactile feedback unit to execute the corresponding vibration mode based on the injury early warning signal; the biomechanical parameters are selected from at least one of the following: joint angle, radius of curvature, and inter-joint coordination ratio derived from multiple joint angles.

2. The smart sportswear based on motion-sensing interaction according to claim 1, characterized in that, The first region includes: The first sub-region corresponds to the midline of the human spine, the peak points of the left and right shoulder joints, and the outer sides of the left and right elbow joints. The second sub-region corresponds to the seventh cervical vertebra at the back of the neck and the distal ends of both arms. The third sub-region corresponds to the pectoralis major, deltoid, and rectus abdominis muscles in the human body; The multimodal sensing unit includes: Multiple flexible strain sensors are distributed in the first sub-region; multiple inertial measurement units are distributed in the second sub-region; and multiple electromyography sensors are distributed in the third sub-region.

3. The smart sportswear based on motion-sensing interaction according to claim 2, characterized in that, The conductive layer of the flexible strain sensor comprises carbon nanotubes and polydimethylsiloxane, wherein the carbon nanotubes are dispersed within the polydimethylsiloxane and form a percolation network; the conductive layer has a microcrack structure. The inertial measurement unit integrates a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer; the inertial measurement unit is configured to operate at a sampling frequency of 500-1000Hz. The electromyography sensor includes a flexible dry electrode array; one end of the flexible dry electrode array is connected to the sensing layer, and the other end passes through the sweat-wicking layer for contact with human skin; the surface of the electrodes in the flexible dry electrode array has micropores with a preset aperture.

4. The smart sportswear based on motion-sensing interaction according to claim 1, characterized in that, The second region corresponds to the sides of the human spine, the shoulder area, and the elbow area; The haptic feedback unit includes: Multiple linear resonant actuators are distributed in the second region; The biomechanical parameters include the joint angle, the radius of curvature, and the inter-joint coordination ratio; the central processing unit is configured to: When the inter-joint coordination ratio is greater than a first preset ratio threshold and less than or equal to a second preset ratio threshold, a motion rhythm deviation signal is generated, and the tactile feedback unit is controlled to execute a low-frequency continuous vibration mode based on the motion rhythm deviation signal. When the joint angle is greater than the first preset joint threshold and less than or equal to the second preset joint threshold, or when the radius of curvature is greater than the first preset curvature threshold and less than or equal to the second preset curvature threshold, a posture deviation signal is generated, and the tactile feedback unit is controlled to execute a high-frequency intermittent pulse vibration mode based on the posture deviation signal. When the inter-joint coordination ratio is greater than the second preset ratio threshold, or the joint angle is greater than the second preset joint threshold, or the radius of curvature is less than or equal to the first preset curvature threshold, an emergency braking signal is generated, and the haptic feedback unit is controlled to execute a full-array burst vibration mode based on the emergency braking signal. The emergency braking signal has a higher generation priority than the action rhythm deviation signal and the attitude deviation signal.

5. The smart sportswear based on motion-sensing interaction according to claim 1, characterized in that, The central processing unit also integrates: The signal preprocessing module is configured to preprocess the multimodal sensing signal, the preprocessing including filtering, time synchronization, zero-point drift elimination and feature extraction of different modal signals; The early warning determination module is connected to the signal preprocessing module and is configured to calculate the biomechanical parameters of the target anatomical site based on the preprocessed signal. When the biomechanical parameters deviate from the preset safety range, it is determined that there is a risk of posture damage and the damage early warning signal is generated. The feedback drive module, connected to the early warning determination module, is configured to receive the damage early warning signal and convert it into a corresponding hardware drive instruction, which is then output to the tactile feedback unit to trigger the corresponding vibration mode.

6. A motion-sensing interaction-based sports injury early warning method, applied to the smart sportswear described in any one of claims 1-5, characterized in that, include: S1 acquires multimodal sensing signals, which include body surface deformation signals, spatial pose signals, and electromyographic signals. S2 calculates the biomechanical parameters of the target anatomical site based on the multimodal sensing signals; When the biomechanical parameters deviate from the preset safety range, a risk of posture injury is determined, and an injury warning signal is generated; wherein, the biomechanical parameters include at least one of joint angle, radius of curvature, and inter-joint coordination ratio derived from multiple joint angles; Based on the damage warning signal, S3 controls the tactile feedback unit located in the tactile sensitive area of ​​the main body of the garment to execute the corresponding vibration mode.

7. The early warning method according to claim 6, characterized in that, The procedure preceding step S1 also includes: Acquire the baseline sensing signal when performing standard actions; A baseline motion profile is constructed based on the baseline sensing signal, and the baseline motion profile includes the baseline values ​​of the biomechanical parameters of the target anatomical site. Step S2 includes: S201 extracts real-time biomechanical parameters at the target anatomical site from the multimodal sensing signal; S202 performs a matching analysis between the real-time biomechanical parameters and the corresponding baseline values ​​in the baseline motion profile; If the real-time biomechanical parameters deviate from the baseline value beyond the preset safety range, it is determined that there is a risk of posture damage, and a damage warning signal is generated.

8. The early warning method according to claim 6, characterized in that, The step of determining a risk of posture damage and generating a damage warning signal when the biomechanical parameters deviate from the preset safety range includes: The preset safety range is determined based on the biomechanical load assessment mechanism; When the inter-joint coordination ratio is greater than the first preset ratio threshold and less than or equal to the second preset ratio threshold, a motion rhythm deviation signal is generated. When the joint angle is greater than the first preset joint threshold and less than or equal to the second preset joint threshold, or when the radius of curvature is greater than the first preset curvature threshold and less than or equal to the second preset curvature threshold, an attitude deviation signal is generated. An emergency braking signal is generated when the inter-joint coordination ratio is greater than the second preset ratio threshold, or the joint angle is greater than the second preset joint threshold, or the radius of curvature is less than or equal to the first preset curvature threshold. The emergency braking signal has a higher generation priority than the action rhythm deviation signal and the attitude deviation signal.

9. The early warning method according to claim 6, characterized in that, Step S2 includes: S211 acquires the multimodal sensing signals within multiple consecutive motion execution cycles and performs temporal feature extraction; S212 constructs a time-accumulated change curve of attitude deviation based on the aforementioned time-series characteristics; S213 calculates the slope characteristics of the time-series cumulative change curve; If the current slope characteristic exceeds the preset slope warning threshold, it is determined that there is a risk of exceeding the safety limit in the next motion execution cycle, and the damage warning signal is generated in advance.

10. The early warning method according to claim 6, characterized in that, The method also includes a tactile teaching mode: Obtain the biomechanical force exertion sequence corresponding to each of the target anatomical sites in the pre-stored standard action model; According to the force exertion sequence, the tactile feedback units distributed at each of the target anatomical sites are activated sequentially to form a temporal vibration wave packet transmitted along the kinetic chain, so as to simulate the muscle force transmission path.

Images