Flexible piezoresistive strain sensor, preparation method of flexible piezoresistive strain sensor, action recognition method and action recognition system

Abstract

The invention discloses a flexible piezoresistive strain sensor and a preparation method thereof, and an action recognition method and system, the flexible piezoresistive strain sensor takes a flexible polymer material as a matrix, and a plurality of conductive fillers are compositely arranged in the matrix, so that the conductive fillers form a continuous composite conductive network structure in a dispersion manner; therefore, stable resistance change is generated under the action of external strain. Based on human body joint strain signals collected by the sensors, resistance signals are preprocessed, and a multi-layer perceptron neural network model is constructed, so that human body actions of different joints and different bending angles are classified and recognized. The structure stability of the flexible piezoresistive strain sensor and the realizability of action recognition are both considered, and the method is suitable for application scenes such as human motion monitoring, rehabilitation evaluation and human-computer interaction.

Description

Technical Field

[0001] This invention relates to the field of flexible electronics and sensor technology, specifically to flexible piezoresistive strain sensors and their fabrication methods, motion recognition methods and systems. Background Technology

[0002] Existing flexible piezoresistive strain sensors typically use elastic polymer materials as the matrix and construct a conductive network by doping with carbon-based or metallic conductive fillers to achieve the conversion between strain and resistance changes. However, significant shortcomings still exist in practical applications.

[0003] First, existing sensors mostly employ a single conductive filler system, which has limited efficiency in reconstructing the conductive path during strain loading and unloading, resulting in long response and recovery times, making it difficult to meet the real-time monitoring requirements of rapid and continuous human movements. Simultaneously, under repeated stretching and bending conditions, the conductive filler is prone to agglomeration, slippage, or localized fracture, leading to a gradual decrease in the stability of the conductive network, significant resistance signal drift, and insufficient reliability for repeated use.

[0004] Furthermore, while improving conductivity, existing material systems often sacrifice flexibility or structural stability, making it difficult to maintain both low resistivity and long-term mechanical reliability within a high strain range. Meanwhile, current technologies primarily focus on the strain signal acquisition stage, and the processing of sensor signals typically relies on simple discrimination or linear analysis, lacking effective mechanisms for recognizing complex multi-joint motion characteristics, thus hindering the achievement of refined and intelligent human motion recognition.

[0005] In view of this, the present invention proposes a flexible piezoresistive strain sensor and its preparation method, motion recognition method and system. Summary of the Invention

[0006] The purpose of this invention is to provide a flexible piezoresistive strain sensor and its preparation method, motion recognition method and system, which aims to solve the problems of insufficient response and cyclic stability of existing flexible piezoresistive strain sensors in rapid dynamic strain monitoring, and lack of effective intelligent recognition of complex human motion signals.

[0007] In a first aspect, the present invention provides a high-performance flexible piezoresistive strain sensor for performing the first aspect, comprising a sensor substrate, wherein the sensor substrate is a PDMS substrate, and at least two types of conductive fillers are compositely disposed inside the sensor substrate, the conductive fillers including a first conductive filler and a second conductive filler; wherein:

[0008] The second conductive filler is a high-entropy alloy powder, the alloy composition of which is CoCrFeMnNi, and the mass ratio of the PDMS matrix to the high-entropy alloy powder is within a preset range.

[0009] The first conductive filler is a multi-walled carbon nanotube. The mass ratio of the PDMS matrix to the multi-walled carbon nanotube is within a preset range, so that the multi-walled carbon nanotube forms a continuous conductive framework. High-entropy alloy powder is distributed between the conductive framework as discrete conductive nodes.

[0010] Under external tensile or bending strain, the continuous conductive skeleton deforms and drives the contact state between the discrete conductive fillers to change, thereby forming a reconfigurable composite conductive network structure inside the sensor substrate, causing the sensor substrate to produce a resistance change corresponding to the strain state.

[0011] As a preferred embodiment of the first aspect of the present invention, the sensor matrix further comprises a silane coupling agent and a diffusion oil; one end of the silane coupling agent is coupled to the hydroxyl groups on the surface of the conductive filler to inhibit the shedding or agglomeration of the conductive filler during repeated strain, and the other end is cross-linked with the molecular chains of the PDMS matrix to form a chemical bonding interface between the conductive filler and the matrix, thereby improving the interfacial compatibility of the composite system and the film formation stability during high strain cycling.

[0012] As a preferred technical solution of the first aspect of the present invention, during the strain loading and release process, the contact state between the continuous conductive frame and the discrete conductive nodes undergoes a reversible change, so that the sensor maintains a stable resistance response under repeated strain.

[0013] In a second aspect, the present invention provides a method for fabricating a high-performance flexible piezoresistive strain sensor, used to fabricate the first aspect, comprising the following steps:

[0014] The PDMS matrix was pretreated by adding diffusion oil and silane coupling agent to the PDMS prepolymer and stirring to ensure that the components were uniformly dispersed in the PDMS matrix.

[0015] Based on the mass of the pretreated PDMS substrate, multi-walled carbon nanotubes and high-entropy alloy powder are added according to a preset mass ratio. The multi-walled carbon nanotubes and the high-entropy alloy powder are added to the PDMS substrate and stirred continuously, so that the multi-walled carbon nanotubes form a continuous conductive framework in the PDMS substrate, and the high-entropy alloy powder is distributed between the conductive framework as discrete conductive nodes.

[0016] Based on the quality of the PDMS substrate, PDMS curing agent is added, stirred evenly, and then poured into a mold of a preset size. The mixture is then heated and cured in an oven under vacuum. After complete curing, the film is demolded to obtain a flexible composite film.

[0017] As a preferred technical solution of the second aspect of the present invention, multi-walled carbon nanotubes are added at a mass ratio of 1:20 to 1:16 based on the mass of the pretreated PDMS substrate, and high-entropy alloy powder is added at a mass ratio of 1:4 to 1:1.

[0018] Thirdly, this invention provides a method for intelligent human motion recognition using a high-performance flexible piezoresistive strain sensor, the application of which includes the following steps:

[0019] A flexible piezoresistive strain sensor is attached to the surface of a human joint, and the flexible piezoresistive strain sensor is arranged along the main bending direction of the corresponding joint. During the bending and extension of the joint, the resistance signal is collected and converted into a time series of relative resistance change rate. The time series of relative resistance change rate is normalized and denoised to form input feature data for human motion recognition.

[0020] A multilayer perceptron neural network model is constructed, comprising a feature input layer, several fully connected layers, several random deactivation layers, and an output layer; the dataset consisting of the input feature data is used as training data to train the multilayer perceptron neural network model.

[0021] The real-time input feature data to be identified is input into the trained multilayer perceptron neural network model, which outputs the classification results corresponding to human joints and joint bending angles, thereby realizing intelligent recognition of human movements.

[0022] As a preferred technical solution of the third aspect of the present invention, the multi-dimensional feature information includes: peak resistance and valley resistance extracted within a single joint bending cycle.

[0023] As a preferred embodiment of the third aspect of the present invention, the specific structure of the multilayer perceptron neural network model is as follows:

[0024] The fully connected layer consists of three layers, with the number of neurons decreasing sequentially according to the layer, and uses the ReLU activation function;

[0025] The random deactivation layer is disposed between adjacent fully connected layers;

[0026] The output layer uses the Softmax activation function to classify action categories.

[0027] As a preferred technical solution of the third aspect of the present invention, the number of classification categories of the output layer corresponds to the human joint types involved in the identification and the bending angle level corresponding to each joint, and is preset in the model building stage.

[0028] Fourthly, the present invention provides a human motion intelligent recognition system based on a high-performance flexible piezoresistive strain sensor, used to perform the third aspect, including:

[0029] The sensing and acquisition unit includes at least one flexible piezoresistive strain sensor, which is attached to the surface of a human joint and arranged along the main bending direction of the corresponding joint. When the human joint undergoes bending or extension movements, it outputs a resistance signal that changes with the joint strain.

[0030] The signal conditioning unit is electrically connected to the sensing and acquisition unit. The signal conditioning unit includes a resistance-to-electrical signal conversion circuit, a filtering circuit, and an analog-to-digital conversion module, which is used to convert, filter, and digitize the resistance signal to obtain a digital strain signal.

[0031] A data processing and feature construction unit is communicatively connected to the signal conditioning unit. The data processing and feature construction unit includes a processor and a memory. The processor is configured to normalize and denoise the digital strain signal and construct feature information for human motion recognition from the processed signal.

[0032] The action recognition unit is communicatively connected to the data processing and feature construction unit. The action recognition unit includes a pre-trained multilayer perceptron neural network model. The multilayer perceptron neural network model includes a feature input layer, at least three fully connected layers, a random deactivation layer set between adjacent fully connected layers, and an output layer. The multilayer perceptron neural network model performs classification operations on the feature information and outputs the corresponding classification results of human joints and joint bending angles.

[0033] The result output unit is connected to the action recognition unit and is used to output the human action recognition result;

[0034] The flexible piezoresistive strain sensor is the high-performance flexible piezoresistive strain sensor described in the first aspect.

[0035] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0036] This invention improves the signal stability of the flexible piezoresistive strain sensor under repeated strain loading conditions by incorporating multiple conductive fillers into a flexible polymer matrix and dispersing these fillers to form a continuous composite conductive network structure. This allows the conductive pathway to exhibit stable and repeatable resistance changes with matrix deformation when the sensor is subjected to tensile or bending strain. Furthermore, based on the strain electrical signals acquired by the flexible piezoresistive strain sensor, a multilayer perceptron neural network model is constructed to process and classify the signals, enabling effective differentiation of strain characteristics corresponding to different joints and bending angles. Thus, this invention balances sensor flexibility and structural integrity with the stability of strain signal acquisition and the feasibility of motion recognition, making it suitable for multi-joint human motion monitoring and recognition applications. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0038] Figure 1 This is a schematic diagram of the overall structure of the flexible piezoresistive strain sensor of the present invention. Figure 1 ;

[0039] Figure 2 This is a schematic diagram of the overall structure of the flexible piezoresistive strain sensor of the present invention. Figure 2 ;

[0040] Figure 3 This is a schematic diagram illustrating the application of the flexible piezoresistive strain sensor of the present invention in the human wrist area;

[0041] Figure 4 This is a schematic diagram illustrating the application of the flexible piezoresistive strain sensor of the present invention in the human elbow area;

[0042] Figure 5 This is a schematic diagram illustrating the application of the flexible piezoresistive strain sensor of the present invention in the human knee area;

[0043] Figure 6 This is a schematic diagram of the intelligent recognition model structure based on MLP of the present invention;

[0044] Figure 7 This is a schematic diagram of the confusion matrix of the classification results of the MLP model of the present invention;

[0045] Figure 8 This is a framework diagram of the human motion intelligent recognition system of the present invention;

[0046] In the figure: 100, flexible piezoresistive strain sensor; 101, PDMS matrix; 102, high-entropy alloy powder; 103, multi-walled carbon nanotubes. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings.

[0048] Throughout the accompanying drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. The described embodiments are only a part of the embodiments of this application, not all of them. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0049] Example 1

[0050] Please see Figure 1-2 As shown, this embodiment provides a high-performance flexible piezoresistive strain sensor. The flexible piezoresistive strain sensor 100 includes a sensor substrate. At least two types of conductive fillers are compositely disposed inside the sensor substrate. The conductive fillers include a first conductive filler and a second conductive filler. The conductive fillers are embedded in the sensor substrate in a dispersed manner to form a continuous composite conductive network structure.

[0051] Specifically, the sensor substrate is a polydimethylsiloxane (PDMS, SYLGARD184) substrate, referred to as PDMS substrate 101. Multi-walled carbon nanotubes 103 (MWCNTs) and high-entropy alloy powder 102 (HEA, CoCrFeMnNi) are synergistically doped in the PDMS substrate 101 to construct a three-dimensional conductive network that achieves an ultrafast response time (approximately 2ms), a fast recovery time (approximately 8ms), low resistivity (approximately 0.739Ω·m), and stability of more than 11,000 cycles.

[0052] A method for fabricating a flexible piezoresistive strain sensor 100 is further provided, comprising the following steps:

[0053] A101. Pretreatment of matrix material: Polydimethylsiloxane (PDMS) prepolymer is selected as the sensor matrix material. Diffusion oil and silane coupling agent are added to the PDMS prepolymer according to a predetermined mass fraction, and the mixture is stirred at room temperature to make the added components uniformly dispersed in the PDMS matrix 101, so as to improve the interfacial compatibility and film formation stability of the composite system.

[0054] A102, conductive filler composite: In the pretreated PDMS matrix 101, multi-walled carbon nanotubes 103 and high-entropy alloy powder 102 are added sequentially. The addition ratio of conductive filler is controlled and the mixture is continuously stirred. The stirring method includes at least one of magnetic stirring, planetary stirring or ultrasonic dispersion. The multi-walled carbon nanotubes 103 and high-entropy alloy powder 102 form an interwoven composite dispersion structure in the PDMS matrix 101, thereby constructing a continuous conductive path.

[0055] A103. Curing and molding: Add PDMS curing agent to the above mixture and continue stirring until the system is uniform. Pour the mixture into a mold of a preset size and then place it in a vacuum drying environment for heating and curing. After complete curing, demold to obtain a flexible composite film.

[0056] The flexible piezoresistive strain sensor 100 prepared by the above steps has good flexibility and stretchability, and its internal conductive network can generate a stable resistance change under external strain.

[0057] In other words, the sensor substrate is polydimethylsiloxane (PDMS). 2 wt.% TSF-96-1000 diffusion oil and 3 wt.% KH-570 silane coupling agent are added to the PDMS prepolymer by mass fraction to improve filler dispersibility and interfacial bonding performance. Subsequently, multi-walled carbon nanotubes 103 are added to the PDMS prepolymer at a mass ratio of 1:20 to 1:16, and high-entropy alloy powder 102 is added at a mass ratio of 1:4 to 1:1. Finally, a PDMS curing agent is added, with a PDMS prepolymer to curing agent mass ratio of 1:15. The above components are magnetically stirred at room temperature for 10 minutes to mix thoroughly. The mixture is poured into a mold and dried in a vacuum drying oven at 80°C for 1 hour. After demolding, a PDMS / HEA / MWCNT flexible composite film is obtained. The mold dimensions are 4cm × 1cm × 1mm. The ratio described in this embodiment is a preferred proportion, which can construct a conductive network in which a continuous conductive framework and discrete conductive nodes work together, thereby achieving good conductivity, strain response characteristics, and cycle stability of the flexible composite film, while maintaining the flexibility and mechanical stability of the material. This embodiment is only a preferred implementation method. When the proportion of the conductive filler varies within a predetermined range, a continuous composite conductive network can still be formed and a stable strain response can be obtained.

[0058] It should also be noted that the addition ratio of multi-walled carbon nanotubes 103 and high-entropy alloy powder 102 can be adjusted within a certain range, the film thickness can be adjusted within the range of 0.8–1.2 mm, and the drying temperature can be adjusted within the range of 70–90 °C to optimize mechanical and electrical properties.

[0059] In this embodiment, the conductive network of the flexible piezoresistive strain sensor 100 is not simply formed by random contact of conductive fillers, but rather is composed of a continuous conductive framework made of multi-walled carbon nanotubes 103 and discrete conductive nodes distributed therebetween. The continuous conductive framework is distributed in a network within the PDMS matrix 101, providing basic conductive pathways; the discrete conductive nodes fill the spaces between the continuous conductive framework, participating in the local connection and modulation of the conductive pathways.

[0060] During strain loading, the sensor substrate undergoes tensile deformation, and the continuous conductive skeleton adjusts its orientation and spacing along with the substrate, causing changes in its contact relationship with adjacent discrete conductive nodes. Some contact connections are weakened or temporarily disconnected. During strain release, as the substrate rebounds, the continuous conductive skeleton gradually restores its original configuration and re-establishes contact connections with the discrete conductive nodes.

[0061] Therefore, the contact connection between the continuous conductive framework and the discrete conductive nodes exhibits a reversible change between the strain loading and release phases. This allows the conductive path to recover through dynamic reconstruction of the contact relationship during repeated strain cycles, rather than undergoing irreversible damage. This recoverable connection mechanism of the conductive path helps maintain a stable conductivity state under multiple strain loading and release conditions, thereby improving the stability and repeatability of the sensor's resistance response.

[0062] Compared to conductive networks that rely solely on a single conductive filler or a simple mixture of multiple fillers, this invention enables the conductive pathway to be dynamically adjusted during deformation through the synergistic effect of a continuous conductive framework and discrete conductive nodes during the strain process, thus avoiding the deterioration of the conductive pathway caused by irreversible shifts in the relative positions of the fillers.

[0063] Furthermore, the flexible piezoresistive strain sensor 100 is subjected to performance tests, including mechanical performance tests, electrical response tests, and cyclic stability tests.

[0064] B101. Mechanical performance test: Fix the sensor film onto the tensile testing device, perform uniaxial tension on the sensor under the preset tension rate, and record the changes in tensile stress and strain to obtain the tensile strength and maximum strain range of the sensor.

[0065] B102. Electrical response test: Under different tensile conditions of the sensor, its resistance signal is collected in real time, and the relative resistance change rate ΔR / R0 is calculated to characterize the response characteristics of the sensor during strain loading and unloading.

[0066] B103. Cyclic stability test: Under the set cyclic stretching frequency and strain amplitude conditions, the sensor is subjected to multiple stretch-release cycle tests, and its resistance change curve is recorded to evaluate the signal stability of the sensor under repeated strain.

[0067] In other words, the flexible composite film has a tensile strength of about 1.968 MPa, a maximum strain of about 279.86%, and after more than 11,000 tensile-unloading cycle tests at a 1 Hz cycle frequency, the resistance changes by about 30%, and the signal still remains stable. The film response time is about 2 ms, the recovery time is about 8 ms, and the resistivity is about 0.739 Ω·m.

[0068] Example 2

[0069] This embodiment, based on the flexible piezoresistive strain sensor 100 prepared in Embodiment 1, attaches the flexible piezoresistive strain sensor 100 to joints such as fingers, wrists, elbows, and knees. It collects resistance signals, preprocesses them, and then inputs them into a multilayer perceptron (MLP) model for intelligent recognition. The MLP model includes: a feature input layer, three fully connected layers (with 512, 256, and 128 neurons respectively), three randomly deactivated layers (with a random deactivation rate of 0.3), and an output layer. The dataset is randomly divided into an 80% training set and a 20% test set to achieve high-precision recognition of the bending angles of fingers, wrists, elbows, and knees. Figure 3-5 Flexible piezoresistive strain sensors 100 were attached to the wrist, elbow, and knee of the human body to illustrate the signal characteristics of the relative resistance change rate over time collected under different bending angles of the wrist, elbow, and knee.

[0070] Specifically, the sensor substrate mounting method includes the following steps:

[0071] C101. Sensor arrangement: The flexible piezoresistive strain sensor 100 is fixed to the corresponding joint surface, and the sensor is arranged along the main bending direction of the joint to ensure that the sensor generates detectable tensile strain when the joint bends.

[0072] C102. Motion data acquisition: Guide the subject to perform periodic bending and stretching movements of the joints at preset angles, collect the resistance signals of different joints under different bending angle conditions, and convert the collected resistance data into a time series of relative resistance change rates.

[0073] The dataset is randomly divided into 80% training set and 20% test set. The activation function of the fully connected layer is ReLU, and the activation function of the output layer is Softmax, so as to achieve high-precision classification of the bending angle of each joint.

[0074] The method for intelligent human motion recognition based on MLP model includes the following steps:

[0075] S101. Signal preprocessing: Normalize, denoise, and process the acquired relative resistance change rate signal to form input feature data for model training and recognition.

[0076] For example, the input feature data includes: peak resistance and valley resistance extracted within a single bending cycle, and corresponding signal category information; the peak resistance and valley resistance are used to characterize the amplitude difference of the resistance response under different joints and different bending angles, and the signal category is used to indicate the joint type or posture state to be identified.

[0077] S102. Model Construction: Construct a multilayer perceptron (MLP) neural network model. The model includes an input layer, several fully connected layers, and an output layer. The output layer is used to output classification results corresponding to different joints and different bending angles, such as... Figure 6 As shown.

[0078] S103. Model training and recognition: The processed dataset is divided into a training set and a test set. The MLP model is trained using the training set, and the classification results of the model are verified based on the test set, thereby realizing the automatic recognition of different joint postures of the human body.

[0079] This can be understood as follows: the number of categories in the output layer is pre-set based on the joint types involved in the identification and the bending angle level corresponding to each joint, used to achieve joint classification of multi-joint and multi-angle posture states. Since the strain signals corresponding to different joints and different bending angles differ in amplitude characteristics such as peak resistance and valley resistance, and these differences are not simply linearly separable, using a multilayer perceptron model to model these features is beneficial for characterizing the nonlinear mapping relationship between different posture states, thereby improving the accuracy of posture classification and avoiding the limitations of fixed threshold or linear discrimination methods in multi-class recognition tasks.

[0080] Through the above implementation methods, the constructed intelligent recognition model can distinguish the strain signals of different joints under different bending angles. The recognition results of each category are mainly distributed on the diagonal position in the confusion matrix, such as... Figure 7The diagram shows a human posture classification confusion matrix based on the signals from the flexible piezoresistive strain sensor. The horizontal axis represents the predicted posture label, and the vertical axis represents the true posture label. The posture labels include various posture states of the elbow, wrist, and knee at different bending angles. This indicates that the model has a good ability to distinguish different action categories, and each true posture is accurately classified into its corresponding predicted category. The classification accuracy of the diagonal elements is 100%, while the off-diagonal elements are 0, indicating no posture recognition confusion. This result shows that the output signal of the flexible piezoresistive strain sensor has significant differences under different joints and different bending angles, which is beneficial for subsequent posture recognition and classification. The human posture recognition method built based on the sensor can achieve stable and accurate classification output under multiple posture conditions, verifying the effectiveness of the sensing system in human motion monitoring and posture recognition applications. This high classification accuracy is mainly attributed to the significant differences in the resistance response generated by the flexible piezoresistive strain sensor under different strain states, which makes the feature signals corresponding to different postures have good separability in the feature space.

[0081] Example 3

[0082] This embodiment, based on the high-performance flexible piezoresistive strain sensor 100 described in Embodiment 1 and the intelligent human motion recognition method described in Embodiment 2, further provides a complete system implementation scheme for continuous monitoring and recognition of multi-joint human motion. The strain signals of human joints collected by the flexible piezoresistive strain sensor 100 are processed through signal conditioning, feature construction, and neural network inference to achieve automatic recognition of human motion states. This is suitable for applications such as human motion monitoring, rehabilitation training assessment, and human-computer interaction.

[0083] Please see Figure 8 The aforementioned intelligent human motion recognition system includes a sensing and acquisition unit, a signal conditioning unit, a data processing and feature construction unit, a motion recognition unit, and a result output unit. The mechanical strain caused by human joint movements is first converted into a resistance change signal by the sensing and acquisition unit. After electrical signal conversion and digitization by the signal conditioning unit, the signal is transmitted to the data processing and feature construction unit to generate feature information for motion recognition. This feature information is further input to the motion recognition unit for classification calculation, and the result output unit outputs the corresponding human motion recognition result, thus forming a complete processing flow from signal acquisition to recognition output. The above units are connected sequentially, and each unit works collaboratively according to a preset processing order. This allows the human joint strain signal to continuously complete the system-level processing flow from raw signal acquisition to motion recognition result output through acquisition, conditioning, feature construction, and classification recognition, forming a complete processing link from human motion to recognition result output.

[0084] The sensing and acquisition unit is used to acquire sensor electrical signals caused by human body movements. The sensing and acquisition unit includes at least one flexible piezoresistive strain sensor 100. The flexible piezoresistive strain sensor 100 adopts the composite thin film structure corresponding to the PDMS matrix 101 / multi-walled carbon nanotube 103 / high-entropy alloy 102 prepared in Example 1.

[0085] In this embodiment, the flexible piezoresistive strain sensor 100 is fixed to the surface of a target joint on the human body via a medical adhesive layer. The sensor is arranged along the main bending direction of the joint. When the joint bends or extends, the sensor generates tensile strain along with the skin surface. Under the action of the joint movement, the continuous conductive skeleton inside the sensor deforms, driving a change in the contact state between discrete conductive nodes, thereby causing the sensor to output a resistance change signal corresponding to the strain state.

[0086] The signal conditioning unit is electrically connected to the sensing acquisition unit. After conditioning the electrical signal, it is transmitted to the data processing and feature construction unit for subsequent action recognition and analysis. It includes a resistor-to-voltage conversion circuit, an analog low-pass filter circuit, and an analog-to-digital conversion module.

[0087] The resistance-to-voltage conversion circuit is used to convert the resistance change output by the flexible piezoresistive strain sensor 100 into a voltage signal;

[0088] Analog filter circuits are used to suppress high-frequency noise introduced by human body micro-movements and environmental electromagnetic interference;

[0089] The analog-to-digital converter module is used to convert analog voltage signals into digital signals and output them to the subsequent processing unit.

[0090] The data processing and feature construction unit includes a processor and a memory. The processor normalizes the acquired digital signals, performs time-domain denoising on the acquired signals, and extracts multi-dimensional feature information. The multi-dimensional feature information is extracted according to a preset time window. The multi-dimensional feature information includes the peak resistance and valley resistance extracted within a single joint bending cycle, forming an input feature dataset for human motion recognition.

[0091] The action recognition unit is communicatively connected to the data processing and feature construction unit, and is used to perform classification operations on the input feature information and output the corresponding human joint type and joint bending angle category; it includes a pre-trained multilayer perceptron neural network model; the multilayer perceptron neural network model includes:

[0092] The feature input layer receives the input feature dataset;

[0093] The fully connected layer comprises at least three layers, with the number of neurons decreasing sequentially according to the layer, and uses the ReLU activation function;

[0094] The random deactivation layer is disposed between adjacent fully connected layers;

[0095] The output layer uses the Softmax activation function to classify action categories.

[0096] The result output unit is connected to the motion recognition unit and is used to output human motion recognition results. The result output unit can take the form of a display interface or a data interface to output the recognition results as joint type and corresponding bending angle level information for use by motion monitoring, rehabilitation assessment or human-computer interaction systems.

[0097] In summary, in the human motion intelligent recognition system described in this embodiment, the mechanical strain generated by the human joint during continuous bending and extension acts on the flexible piezoresistive strain sensor 100, causing the flexible piezoresistive strain sensor 100 to output a resistance change signal corresponding to the joint strain state. The sensing acquisition unit is used to collect the resistance signal of the flexible piezoresistive strain sensor caused by human motion, thereby achieving stable acquisition of human joint strain signals. The acquired resistance change signal undergoes resistance-to-electrical signal conversion, filtering, and digitization processing by the signal conditioning unit, and the influence of noise interference on motion recognition is suppressed through the signal conditioning and subsequent feature construction process. After conditioning the acquired resistance signal, the signal conditioning unit transmits it to the data processing and feature construction unit for subsequent human motion recognition analysis. The data processing and feature construction unit generates feature information for motion recognition based on the digitized signal. The motion recognition unit further performs classification calculations on the feature information based on a multilayer perceptron neural network model, enabling the differentiation of strain features corresponding to different joints and different bending angles. Finally, the result output unit outputs the corresponding human motion recognition result, thereby achieving system-level feasibility of human motion recognition while ensuring sensor flexibility and structural stability.

[0098] 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 high-performance flexible piezoresistive strain sensor, characterized in that, The sensor includes a sensor substrate, which is a PDMS substrate. At least two types of conductive fillers are compositely disposed within the sensor substrate, including a first conductive filler and a second conductive filler; wherein: The second conductive filler is a high-entropy alloy powder, the alloy composition of which is CoCrFeMnNi, and the mass ratio of the PDMS matrix to the high-entropy alloy powder is within a preset range. The first conductive filler is a multi-walled carbon nanotube. The mass ratio of the PDMS matrix to the multi-walled carbon nanotube is within a preset range, so that the multi-walled carbon nanotube forms a continuous conductive framework. High-entropy alloy powder is distributed between the conductive framework as discrete conductive nodes. Under external tensile or bending strain, the continuous conductive skeleton deforms and drives the contact state between the discrete conductive fillers to change, thereby forming a reconfigurable composite conductive network structure inside the sensor substrate, causing the sensor substrate to produce a resistance change corresponding to the strain state.

2. The high-performance flexible piezoresistive strain sensor according to claim 1, characterized in that, The sensor matrix also contains a silane coupling agent and a diffusion oil. One end of the silane coupling agent is coupled to the hydroxyl groups on the surface of the conductive filler to inhibit the shedding or aggregation of the conductive filler during repeated strain. The other end is cross-linked with the molecular chains of the PDMS matrix to form a chemical bonding interface between the conductive filler and the matrix, thereby improving the interfacial compatibility of the composite system and the film formation stability during high strain cycling.

3. The high-performance flexible piezoresistive strain sensor according to claim 2, characterized in that, During the strain loading and release process, as the continuous conductive skeleton deforms with the sensor substrate, the contact connection state between it and the discrete conductive nodes distributed therein reversibly switches between the loading and release phases. This allows the conductive path to maintain a recoverable connection during repeated strain cycles, thereby enabling the sensor to output a stable resistance response under multiple strain loading and release conditions.

4. A method for fabricating a high-performance flexible piezoresistive strain sensor, used to fabricate the high-performance flexible piezoresistive strain sensor according to any one of claims 1-3, characterized in that, Includes the following steps: The PDMS matrix was pretreated by adding diffusion oil and silane coupling agent to the PDMS prepolymer and stirring to ensure that the components were uniformly dispersed in the PDMS matrix. Based on the mass of the pretreated PDMS substrate, multi-walled carbon nanotubes and high-entropy alloy powder are added according to a preset mass ratio. The multi-walled carbon nanotubes and the high-entropy alloy powder are added to the PDMS substrate and stirred continuously, so that the multi-walled carbon nanotubes form a continuous conductive framework in the PDMS substrate, and the high-entropy alloy powder is distributed between the conductive framework as discrete conductive nodes. Based on the quality of the PDMS substrate, PDMS curing agent is added, stirred evenly, and then poured into a mold of a preset size. The mixture is then heated and cured in an oven under vacuum. After complete curing, the film is demolded to obtain a flexible composite film.

5. The method for fabricating a high-performance flexible piezoresistive strain sensor according to claim 4, characterized in that, Based on the mass of the pretreated PDMS substrate, multi-walled carbon nanotubes were added at a mass ratio of 1:20 to 1:16, and high-entropy alloy powder was added at a mass ratio of 1:4 to 1:

1.

6. A method for intelligent human motion recognition using a high-performance flexible piezoresistive strain sensor, based on the application of a high-performance flexible piezoresistive strain sensor as described in any one of claims 1-3, characterized in that, Includes the following steps: A flexible piezoresistive strain sensor is attached to the surface of a human joint, and the flexible piezoresistive strain sensor is arranged along the main bending direction of the corresponding joint. During the bending and extension of the joint, the resistance signal is collected and converted into a time series of relative resistance change rate. The time series of relative resistance change rate is normalized and denoised to extract multi-dimensional feature information and form an input feature dataset for human motion recognition. A multilayer perceptron neural network model is constructed, comprising a feature input layer, several fully connected layers, several random deactivation layers, and an output layer; the input feature dataset is used as training data to train the multilayer perceptron neural network model. The multi-dimensional feature information to be identified is input into the trained multilayer perceptron neural network model, and the corresponding classification results of human joints and joint bending angles are output, thereby realizing intelligent recognition of human movements.

7. The method for intelligent human motion recognition using a high-performance flexible piezoresistive strain sensor according to claim 6, characterized in that, The multi-dimensional feature information includes: peak resistance and valley resistance extracted within a single joint bending cycle.

8. The method for intelligent human motion recognition using a high-performance flexible piezoresistive strain sensor according to claim 7, characterized in that, The specific structure of the multilayer perceptron neural network model is as follows: The fully connected layer comprises at least three layers, with the number of neurons decreasing sequentially according to the layer, and uses the ReLU activation function; The random deactivation layer is disposed between adjacent fully connected layers; The output layer uses the Softmax activation function to classify action categories.

9. The method for intelligent human motion recognition using a high-performance flexible piezoresistive strain sensor according to claim 8, characterized in that, The number of classification categories in the output layer corresponds to the types of human joints involved in the identification and the bending angle level of each joint, and is preset during the model building stage.

10. A human motion intelligent recognition system based on a high-performance flexible piezoresistive strain sensor, characterized in that, It includes a sensor acquisition unit, a signal conditioning unit, a data processing and feature construction unit, an action recognition unit, and a result output unit, wherein: The sensing and acquisition unit includes at least one flexible piezoresistive strain sensor, which is attached to the surface of a human joint and arranged along the main bending direction of the corresponding joint. When the human joint undergoes bending or extension movements, it outputs a resistance signal that changes with the joint strain. The signal conditioning unit is electrically connected to the sensing and acquisition unit. The signal conditioning unit includes a resistance-to-electrical signal conversion circuit, a filtering circuit, and an analog-to-digital conversion module, which is used to convert the resistance signal into a digital strain signal. A data processing and feature construction unit is communicatively connected to the signal conditioning unit. The data processing and feature construction unit includes a processor and a memory. The processor is configured to normalize and denoise the digital strain signal and construct feature information for human motion recognition from the processed signal. The action recognition unit is communicatively connected to the data processing and feature construction unit. The action recognition unit includes a pre-trained multilayer perceptron neural network model. The multilayer perceptron neural network model includes a feature input layer, at least three fully connected layers, a random deactivation layer set between adjacent fully connected layers, and an output layer. The multilayer perceptron neural network model outputs the corresponding classification results of human joints and joint bending angles based on the feature information. The result output unit is connected to the action recognition unit and is used to output the human action recognition result; The flexible piezoresistive strain sensor is the high-performance flexible piezoresistive strain sensor described in any one of claims 1–3.

Images