High-density electromyographic sensor array, acquisition system, and preparation and gesture recognition methods

Abstract

A high-density electromyographic sensor array, an acquisition system, and preparation and gesture recognition methods. The arrangement of sensors is convenient and flexible, and a high-density 64-lead electromyographic sensor array is adopted, without relying on a gel electrode patch, so that the arrangement of the sensors is more convenient. The high-density electromyographic sensor array is simple and convenient to prepare and suitable for rehabilitation, and adopts specific materials which are simple and convenient to prepare and suitable for rehabilitation application, thereby avoiding the complexity of a PDMS material. Acquisition hardware is cost-effective, so that a high-density electromyographic acquisition hardware system for large-scale application is provided, thereby achieving a low cost, and facilitating secondary development. A flexible wifi communication mode is adopted, as the wifi communication mode is more flexible than a traditional USB mode, thereby facilitating system deployment and data transmission. A gesture recognition algorithm is based on hybrid learning, and the gesture recognition accuracy is improved by means of the hybrid learning algorithm, so that a rehabilitation robot can respond to an action of a user in a more intelligent and personalized manner, thereby improving the rehabilitation effect.

Description

High-density electromyographic sensor array, acquisition system, preparation and gesture recognition method Technical Field

[0001] The present invention relates to the field of intelligent hand rehabilitation technology, and in particular to a high-density myoelectric sensor array, an acquisition system, a preparation method, and a gesture recognition method. Background Art

[0002] Currently, there are several high-density EMG sensors and related technologies on the market. Most utilize multiple gel electrodes, while others are fabricated using PDMS, primarily for scientific research. Additionally, there are some multi-channel EMG acquisition hardware and EMG gesture recognition algorithms. The main issues are:

[0003] The existing high-density electromyographic sensors are not easily arranged. Most sensors use gel electrode patches, which are relatively inconvenient to arrange, especially when used for a long time, and the electrodes need to be replaced frequently.

[0004] High-density electromyoelectric sensors are made of complex PDMS materials. PDMS sensors are mostly used in scientific research, but their preparation and use are too complicated and not suitable for large-scale rehabilitation applications.

[0005] The multi-channel acquisition hardware for electromyography acquisition is large, costly, and difficult to develop, which limits its feasibility in large-scale applications and makes secondary development difficult.

[0006] The transmission interface is limited. The data transmission interface of existing equipment is limited, which affects the real-time and flexibility of data.

[0007] The recognition accuracy needs to be improved. The existing electromyographic gesture recognition algorithm has certain limitations in improving the accuracy of user intention recognition.

[0008] Summary of the Invention

[0009] In order to achieve the above-mentioned objectives and other advantages of the present invention, the first objective of the present invention is to provide a high-density electromyographic sensor array, including a plurality of lead electrodes and conductive channels, the plurality of lead electrodes are divided into a plurality of lead electrode groups, each lead electrode group is arranged at intervals along a first direction, the lead electrodes in each lead electrode group are arranged at intervals along a second direction, the lead electrodes are connected to the conductive channels, and the conductive channels corresponding to each lead electrode are arranged at intervals according to a preset shape.

[0010] Furthermore, the lead electrode group is divided into a first lead electrode group and a second lead electrode group, and the first lead electrode group and the corresponding conductive channel are arranged in a mirror-symmetrical manner with the second lead and the corresponding conductive channel along a midline.

[0011] Furthermore, each first lead electrode group or each second lead electrode group is arranged at equal intervals along the first direction, and the lead electrodes in each lead electrode group are arranged at equal intervals along the second direction.

[0012] Furthermore, the first direction is perpendicular to the second direction.

[0013] Furthermore, the number of the lead electrodes is 64, and the number of the lead electrode groups is 16.

[0014] A second object of the present invention is to provide a method for preparing a high-density myoelectric sensor array, which is used to prepare the above-mentioned high-density myoelectric sensor array, comprising the following steps:

[0015] Prepare PTE substrate, cut or prepare PTE substrate sheets, and ensure that the size of the PTE substrate sheets is suitable for the design of high-density EMG sensor arrays;

[0016] Silver paste preparation: for the design of high-density electromyographic sensor arrays, corresponding silver particles and conductive colloids are formulated to prepare sparse conductive silver paste;

[0017] Screen printing: Using screen printing technology, sparse conductive silver paste is evenly coated on the PTE substrate to form a high-density layout of 64 leads;

[0018] Drying treatment: Drying is carried out at a set temperature to remove the solvent in printing and ensure the stability of the conductive silver paste;

[0019] High temperature curing: placing the printed high-density electromyographic sensor array in a high temperature environment to ensure that the conductive silver paste is firmly bonded to the PTE substrate sheet to form a stable conductive channel;

[0020] Detection and adjustment: The prepared high-density electromyographic sensor array is tested to ensure the connectivity and sparsity of the conductive channels.

[0021] The third object of the present invention is to provide a high-density myoelectric acquisition system for realizing the simultaneous acquisition of the above-mentioned high-density myoelectric sensor array, including several analog front-end chips, main control chips, communication modules, sensor modules, feedback and display modules, driver chips, storage modules, charging and voltage stabilization modules, and clock modules; wherein,

[0022] The analog front-end chip is used to provide high-precision analog signal conversion for each conductive channel in the high-density electromyographic sensor array, realizing simultaneous acquisition of all leads;

[0023] The communication module is used for data communication to realize data transmission;

[0024] The main control chip is used to control the operation of the acquisition system;

[0025] The sensor module is used to obtain posture information about limb movement;

[0026] The clock module is used to synchronize the operation of each module to ensure the accuracy of data collection;

[0027] The feedback and display module is used to provide user feedback or display system status;

[0028] The driver chip is used to control the feedback and display modules;

[0029] The storage module is used to store the collected electromyographic signal data;

[0030] The charging and voltage stabilization module is used to manage the power supply of the acquisition system to ensure the stable operation of each module.

[0031] Furthermore, several of the analog front-end chips are connected to the main control chip via an SPI interface to form a daisy chain.

[0032] Furthermore, the communication module includes a wireless communication module and a USB Type-C interface. The wireless communication module is used for data communication to realize wireless data transmission. The USB Type-C interface is used for data communication to realize data transmission, and is reused for the debug download port of the main control chip, the debug download port of the wireless communication module, the USB bus of the storage module, and the charging pin of the charging and voltage stabilization module.

[0033] Furthermore, the sensor module adopts a 9-axis acceleration sensor; the feedback and display module adopts a vibration motor and an LED display module.

[0034] Furthermore, the data transmission between the high-density electromyography acquisition system and the high-density electromyography sensor array adopts a frame structure, and the frame structure includes a start flag, a sampling sequence number, multiple channel data, a test flag bit and a check bit.

[0035] A fourth object of the present invention is to provide a high-density myoelectric gesture recognition method, which is applied to the above-mentioned high-density myoelectric acquisition system and includes the following steps:

[0036] Real-time collection of electromyographic signal data;

[0037] Perform filtering, denoising and image construction processing on the collected electromyographic signal data;

[0038] The processed electromyographic signal data is input into the electromyographic gesture recognition model to obtain real-time control decisions; the network structure of the electromyographic gesture recognition model includes a one-dimensional convolutional neural network, a two-dimensional convolutional neural network, an alternating convolutional neural network, a multi-stream convolution operation and a large pooling layer.

[0039] Furthermore, the construction of the myoelectric gesture recognition model includes the following steps:

[0040] Preprocessing of electromyographic signal data;

[0041] Perform feature extraction, feature selection, and feature dimensionality reduction on the preprocessed electromyographic signal data;

[0042] The processed features are input into the myoelectric gesture recognition model for training;

[0043] The following steps are also included:

[0044] Smoothing the prediction results, judging multiple times within a preset time, and using multiple votes as control instructions;

[0045] When the rehabilitation robot is moving, it does not accept control instructions;

[0046] When the rehabilitation robot is in the hand-closed state, it only accepts hand-opening instructions.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention provides a high-density electromyographic sensor array, an acquisition system, a preparation method, and a gesture recognition method, which facilitate and flexibly arrange sensors. A high-density 64-lead electromyographic sensor array is used, which does not rely on gel electrode patches, making the arrangement of sensors more convenient. The high-density electromyographic sensor array is easy to prepare and suitable for rehabilitation. It uses specific materials, is easy to prepare, and is suitable for rehabilitation applications, avoiding the complexity of PDMS materials. Cost-effective acquisition hardware provides a high-density electromyographic acquisition hardware system for large-scale applications, which is low in cost and easy for secondary development. A flexible Wi-Fi communication method is used, which is more flexible than the traditional USB method and facilitates system deployment and data transmission. A gesture recognition algorithm based on hybrid learning is used to improve the accuracy of gesture recognition through the hybrid learning algorithm, so that the rehabilitation robot can respond to user actions more intelligently and personalized, thereby improving the rehabilitation effect.

[0049] This invention, through the combination of high-density electromyographic sensors and a hybrid learning algorithm, enables more precise and accurate monitoring and identification of hand muscle activity, providing a higher level of personalization and real-time capabilities for rehabilitation treatment. This allows rehabilitation professionals to more effectively develop and adjust rehabilitation plans, while enabling patients to more autonomously engage in rehabilitation training in their daily lives. This is expected to bring significant advancements to the field of rehabilitation therapy, enhance rehabilitation outcomes, and improve patients' quality of life.

[0050] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and to implement it according to the contents of the description, the following preferred embodiments of the present invention are described in detail with reference to the accompanying drawings. The specific implementation methods of the present invention are given in detail by the following embodiments and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0052] FIG1 is a schematic diagram of a 64-lead electromyographic sensor of Example 1;

[0053] FIG2 is a flow chart of a method for preparing a high-density electromyographic sensor array according to Example 2;

[0054] FIG3 is a schematic diagram of a high-density myoelectric acquisition system according to Example 3;

[0055] FIG4 is a schematic diagram showing a high-density myoelectric acquisition system according to Example 3;

[0056] FIG5 is a flow chart of offline training and online real-time control of Example 4;

[0057] FIG6 is a flow chart 1 of a neural network processing myoelectric signal data according to Example 4;

[0058] FIG7 is a schematic diagram of action label correction based on the maximum area method in Example 4;

[0059] FIG8 is a second flow chart of the neural network processing myoelectric signal data in Example 4.

[0060] In the figure: 1. Lead electrode; 2. Conductive channel. DETAILED DESCRIPTION

[0061] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0062] Example 1

[0063] A high-density myoelectric sensor array is primarily used in the field of rehabilitation medicine, particularly for rehabilitation training for hand dysfunction. As shown in Figure 1, it includes several lead electrodes 1 and conductive channels 2. The lead electrodes are divided into several lead electrode groups, each of which is spaced apart along a first direction. The lead electrodes within each lead electrode group are spaced apart along a second direction. The lead electrodes are connected to the conductive channels, and the conductive channels corresponding to each lead electrode are spaced apart in a predetermined pattern.

[0064] The lead electrode group is divided into a first lead electrode group and a second lead electrode group. The first lead electrode group and the corresponding conductive channel are arranged in a mirror-symmetrical manner with the second lead and the corresponding conductive channel along the midline.

[0065] Each lead electrode group is arranged at equal intervals along a first direction, and the lead electrodes in each lead electrode group are arranged at equal intervals along a second direction, wherein the first direction is perpendicular to the second direction.

[0066] Figure 1 shows a 64-lead electromyographic sensor. The sensor comprises 64 lead electrodes. Four longitudinal lead electrodes 1 form a lead electrode group, resulting in 16 lead electrode groups. The lead electrode groups are divided into eight first lead electrode groups and eight second lead electrode groups. The eight first lead electrode groups and their corresponding conductive channels are arranged in mirror-symmetric fashion along the midline with the eight second leads and their corresponding conductive channels. The first direction is the horizontal direction in Figure 1, and the second direction is the vertical direction in Figure 1. Each first lead electrode group or each second lead electrode group is equally spaced along the first direction, while the lead electrodes within each lead electrode group are equally spaced along the second direction. The lead electrodes are connected to the conductive channels, and the vast majority of the conductive channels are arranged in a zigzag pattern.

[0067] This embodiment adopts a 64-lead high-density EMG sensor array layout, taking into account muscle anatomy and physiology to ensure that the lead position design should fully cover the target muscle area to obtain comprehensive EMG signals; the spacing and arrangement between the leads are optimized to balance the high-density layout and sensor sensitivity.

[0068] Example 2

[0069] A method for preparing a high-density myoelectric sensor array is provided for preparing the above-mentioned high-density myoelectric sensor array. A detailed description of the high-density myoelectric sensor array can be found in the corresponding description of the above-mentioned high-density myoelectric sensor array embodiment, and will not be repeated here. As shown in FIG2 , the preparation method comprises the following steps:

[0070] S1. Prepare PTE substrate, cut or prepare PTE substrate sheets, and ensure that the size of the PTE substrate sheets is suitable for the design of high-density EMG sensor arrays;

[0071] S2. Silver paste preparation: for the design of high-density electromyographic sensor arrays, appropriate amounts of silver particles and conductive colloids are mixed to prepare sparse conductive silver paste;

[0072] S3, screen printing, using screen printing technology, evenly coating sparse conductive silver paste on the PTE substrate to form a high-density layout of 64 leads;

[0073] S4, drying treatment, drying at an appropriate temperature to remove the solvent in printing and ensure the stability of the conductive silver paste;

[0074] S5, high temperature curing, placing the printed high-density electromyographic sensor array in a high temperature environment to ensure that the conductive silver paste is firmly bonded to the PTE substrate sheet to form a stable conductive channel;

[0075] S6. Detection and adjustment: Detect the prepared high-density electromyographic sensor array to ensure the connectivity and sparsity of the conductive channels.

[0076] Adjust and optimize as needed to ensure the sensor performs well in actual applications.

[0077] The high-density electromyographic sensor array preparation method provided in this embodiment can achieve high-density layout: through the 64-lead design, high-density electromyographic signal collection is achieved, allowing the sensor to monitor the activity of the target muscle area more comprehensively and accurately.

[0078] Advantages of PTE material: Using PTE as the base material, it is soft and biocompatible, can better adapt to the skin surface and improve wearing comfort.

[0079] Screen printing technology: Screen printing technology can achieve uniform conductive silver paste coating, ensuring the uniformity and sparseness of the conductive channel to improve the performance of the sensor.

[0080] High temperature curing: The high temperature curing step ensures a strong bond between the conductive silver paste and the PTE substrate, improving the stability and durability of the sensor.

[0081] Suitable for large-scale preparation: The process adopted in this preparation method is relatively simple, suitable for large-scale production, and reduces the preparation cost.

[0082] Flexible conductive channel design: By adjusting the layout of screen-printed conductive silver paste, flexible conductive channels can be designed on the sensor to adapt to different muscle structures and signal requirements.

[0083] The high-density electromyographic sensor array preparation method provided in the present embodiment adopts PTE material as substrate, and is prepared by steps such as screen printing of sparse conductive silver paste and high-temperature curing. The advantage of this method is that it selects soft PTE material to provide comfort, and a high-density layout of 64 leads can be achieved through screen printing technology, so that the electromyographic sensor can accurately and comprehensively capture muscle activity. In addition, high-temperature curing is used to ensure that the conductive silver paste is well combined with the PTE substrate, thereby improving the stability and durability of the sensor. The comprehensive use of soft materials, screen printing technology and high-temperature curing provides an excellent solution for the design and preparation of high-density electromyographic sensors.

[0084] Example 3

[0085] A high-density myoelectric acquisition system is used to achieve simultaneous acquisition of the above-mentioned high-density myoelectric sensor array, which is suitable for large-scale data acquisition. For a detailed description of the high-density myoelectric sensor array, please refer to the corresponding description in the above-mentioned high-density myoelectric sensor array embodiment, which will not be repeated here. As shown in Figure 3, the acquisition system includes several analog front-end chips, a main control chip, a communication module, a sensor module, a feedback and display module, a driver chip, a storage module, a charging and voltage stabilization module, and a clock module; wherein,

[0086] The analog front-end chip provides high-precision analog signal conversion for each conductive channel in the high-density EMG sensor array, enabling simultaneous acquisition of all leads. Specifically, the analog front-end uses an amplifier to amplify and filter the 64-lead bioelectrical signals. Several analog front-end chips are connected to the main control chip via an SPI interface, forming a daisy-chain. This embodiment uses eight 8-channel ADS1299 front-end amplifier chips in a daisy-chain design. The ADS1299 provides high-precision analog signal conversion for each channel, enabling simultaneous acquisition of all 64 leads.

[0087] The communication module is used for data communication and data transmission. Specifically, the communication module includes a wireless communication module and a USB Type-C interface. The wireless communication module is used for data communication and wireless data transmission. The USB Type-C interface is also used for data communication and data transmission. It is also reused for the debug and download port of the main control chip, the debug and download port of the wireless communication module, the USB bus of the storage module, and the charging pins of the charging and voltage regulation module. In this embodiment, the wireless communication module uses the WiFi module USRC 322 for data communication and wireless data transmission.

[0088] The main control chip controls the operation of the acquisition system; an STM32 chip is used for controlling the entire system. The USB Type-C port is also used for debugging and downloading the STM32 chip, ensuring system stability and debuggability. The USR C322 WiFi module communicates with the STM32 chip via the serial port, enabling remote data transmission.

[0089] The sensor module is used to obtain posture information about limb movement. In this embodiment, the sensor module uses a 9-axis acceleration sensor. The 9-axis acceleration sensor module is connected to Stm32 through an I2C interface to provide information about limb movement.

[0090] The clock module is used to synchronize the operation of each module to ensure the accuracy of data collection;

[0091] The feedback and display module is used to provide user feedback or display system status. In this embodiment, the feedback and display module uses a vibration motor and an LED display module, which are controlled by an Stm32 chip to provide user feedback or display system status, as shown in FIG4 .

[0092] The driver chip is used to control the feedback and display modules. This embodiment uses the ULN2003 driver chip to control the vibration motor or other peripherals that need to be driven.

[0093] The storage module is used to store the collected electromyographic signal data; this embodiment only integrates Nand Flash for data storage.

[0094] The charging and voltage stabilization module is used to manage the power supply of the acquisition system to ensure the stable operation of each module.

[0095] The high-density myoelectric acquisition system provided in this embodiment can achieve high-density acquisition: the 64-lead design achieves high-density bioelectric signal acquisition through the ADS1299 chip.

[0096] Wireless communication: WiFi module is used to realize wireless data transmission, which improves the flexibility and portability of the system.

[0097] USB Type-C port multiplexing: Flexible multiplexing of the USB Type-C port, including debugging and downloading, USB bus communication, and charging, provides diverse interface uses.

[0098] Versatility: The system integrates a 9-axis acceleration sensor module, feedback module, storage module, etc., making it more versatile.

[0099] Comprehensive control: The main control chip Stm32 is responsible for the control of the entire system, realizing the coordination of various modules and ensuring accurate data collection and transmission.

[0100] The data transmission between the high-density electromyography acquisition system and the high-density electromyography sensor array adopts a frame structure, which includes a start flag, a sampling sequence number, multiple channel data, a test flag bit and a check bit.

[0101] In this embodiment, one frame of sampled data is 3040 bytes, including data from 10 samplings. One sampling data is 304 bytes, including 8 groups of sampled data, each group has 8 channels, for a total of 64 channels.

[0102] Single sampling example: AA AA F1 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data AA AA F2 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data AA AA F3 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data AA AA F4 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data AA AA F5 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data AA AA F6 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data AA AA F7 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data AA AA F8 21(4*8+1=33)...test CHECK / / 4+32+1+1=38 bytes of 8-channel data

[0103] In hexadecimal notation, each frame of data includes a start flag (AA AA), a sample number (F1 to F8), 32 bytes of 8-channel data, a test flag, and a check digit. This embodiment uses a specific communication protocol, such as AA AA F1 21...CHECK, where CHECK is the check digit. Wireless data transmission is performed via the WiFi module, and wired debugging and data download are performed using the USB Type-C port.

[0104] The data transmission protocol provided in this embodiment can realize high-density data transmission: using a data structure of 3040 bytes per frame, it can realize the effective collection and transmission of 64 channels of high-density electromyographic signals.

[0105] Multiple sampling: Each frame includes data sampled 10 times, which improves the data richness and sampling frequency, and helps to more comprehensively understand muscle activity.

[0106] Data consistency: The consistency and reliability of transmitted data are ensured through designs such as the start flag, sampling sequence number, and check bit.

[0107] Flexible Communication: The WiFi module is used for wireless data transmission, which improves the flexibility and portability of the system. The multifunctional USB Type-C port supports wired debugging, data download, USB bus communication and charging, making the system more versatile.

[0108] Strong debuggability: Using the USB Type-C port as a debug and download port improves the debuggability of the system, making it easier for engineers to debug and maintain the system.

[0109] Suitable for real-time monitoring: Due to the high efficiency of data transmission and the characteristics of multiple sampling, it is suitable for real-time monitoring and analysis of electromyographic signals, providing strong support for rehabilitation applications.

[0110] Overall, this data transmission protocol is efficient, reliable, and flexible in high-density EMG data acquisition and transmission, and is suitable for real-time monitoring and rehabilitation applications.

[0111] Example 4

[0112] A high-density myoelectric gesture recognition method is applied to the above-mentioned high-density myoelectric acquisition system. For a detailed description of the acquisition system, please refer to the corresponding description in the above-mentioned acquisition system embodiment and will not be repeated here. As shown in Figures 5, 6, and 8, the recognition method includes the following steps:

[0113] Real-time collection of electromyographic signal data (sEMG);

[0114] Perform filtering, denoising and image construction processing on the collected electromyographic signal data;

[0115] The processed EMG signal data is input into the EMG gesture recognition model to obtain real-time control decisions. The network structure of the EMG gesture recognition model includes a one-dimensional convolutional neural network (Conv1D), a two-dimensional convolutional neural network (Conv2D), an alternating convolutional neural network (Alternate-CNN), multi-stream convolution operations, and a large pooling layer (ML-CNN).

[0116] This embodiment performs smoothing on the prediction results, and judges multiple times within a preset time, for example, 10 times in 1 second, and uses multiple votes (3-5 times) as control instructions;

[0117] When the rehabilitation robot is moving, it does not accept control instructions;

[0118] When the rehabilitation robot is in the hand-closed state, it only accepts hand-opening instructions.

[0119] The myoelectric gesture recognition model uses the myoelectric signal training data for offline training. The data is filtered and denoised, and the image is constructed and then input into the convolutional neural network (CNN) model for training.

[0120] Specifically, the construction of the myoelectric gesture recognition model includes the following steps:

[0121] The EMG signal data was preprocessed, including 10-350 Hz bandpass filter and 50 Hz notch filter processing; label correction, which can be done by data cropping, maximum area method (as shown in Figure 7), and maximum likelihood correction; sample imbalance problem: processing of resting movements (through thresholding); feature normalization: min-max standardization, standard deviation normalization; data enhancement: adding Gaussian noise, flipping signal channels, and time window + incremental window.

[0122] The preprocessed electromyographic signal data is subjected to feature extraction, feature selection, and feature dimensionality reduction processing; feature extraction includes time domain, frequency domain, and time-frequency domain (tsfresh library); feature selection includes filtering methods: variance selection method, correlation coefficient method, chi-square test, mutual information method, to evaluate the degree of correlation between a single feature and the result value, and sort to retain the top relevant feature parts; wrapping type: recursive feature deletion method, feature sorting based on learning model; embedding type: regularization method (L1 regularization screening features); feature dimensionality reduction includes PCA, LDA, SVD decomposition, popular learning LLE (non-linear dimensionality reduction), autoencoder, and T-SNE.

[0123] The processed features are input into the electromyographic gesture recognition model for training; specifically including KNN, LDA, DT, LR, NB, SVM, ANN; RF, AdaBoost, GBDT, LightGBM, XGBoost; AE, MLP, deep Boltzmann machine, deep belief network, CNN, RNN, LSTM, Inception, Attention; transfer learning, GAN.

[0124] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0125] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0126] The above are merely examples of the present invention and are not intended to limit one or more embodiments of the present invention. For those skilled in the art, various modifications and variations of one or more embodiments of the present invention may be made. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present invention shall be included within the scope of the claims of one or more embodiments of the present invention.

Claims

1. A high-density electromyography sensor array, characterized in that: It includes a number of lead electrodes and conductive channels. The number of lead electrodes is divided into a number of lead electrode groups. Each lead electrode group is arranged at intervals in the first direction, and the lead electrodes in each lead electrode group are arranged at intervals in the second direction. The lead electrodes are connected to the conductive channels, and the conductive channels corresponding to each lead electrode are arranged at intervals in a preset shape.

2. The high-density electromyography sensor array according to claim 1, characterized in that: The lead electrode group is divided into a first lead electrode group and a second lead electrode group. The first lead electrode group and the corresponding conductive channels and the second lead and the corresponding conductive channels are arranged symmetrically about the midline.

3. The high-density electromyography sensor array according to claim 2, characterized in that: Each first lead electrode group or each second lead electrode group is arranged at equal intervals in the first direction, and the lead electrodes in each lead electrode group are arranged at equal intervals in the second direction.

4. The high-density electromyography sensor array according to claim 3, characterized in that: The first direction is perpendicular to the second direction.

5. The high-density electromyography sensor array according to claim 4, characterized in that: The number of lead electrodes is 64, and the number of lead electrode groups is 16.

6. A method for preparing a high-density electromyography sensor array for preparing the high-density electromyography sensor array according to any one of claims 1 to 5, characterized in that, It includes the following steps: Prepare a PTE substrate, cut or prepare a PTE substrate sheet to ensure that the size of the PTE substrate sheet is suitable for the design of the high-density electromyography sensor array; Silver paste preparation, for the design of the high-density electromyography sensor array, prepare the corresponding silver particles and conductive colloid to prepare a sparse conductive silver paste; Screen printing, using screen printing technology, evenly coat the sparse conductive silver paste on the PTE substrate sheet to form a high-density layout of 64 leads; Drying treatment, drying at a set temperature to remove the solvent in the printing and ensure the stability of the conductive silver paste; High-temperature curing, place the printed high-density electromyography sensor array in a high-temperature environment to ensure the firm bonding of the conductive silver paste and the PTE substrate sheet to form a stable conductive channel; Detection and adjustment, detect the prepared high-density electromyography sensor array to ensure the connectivity and sparsity of the conductive channels.

7. A high-density electromyography acquisition system for simultaneously acquiring the high-density electromyography sensor array according to any one of claims 1 to 5, characterized in that: It includes a number of analog front-end chips, a main control chip, a communication module, a sensor module, a feedback and display module, a driver chip, a storage module, a charging and voltage stabilizing module, and a clock module; among them, The analog front-end chip is used to provide high-precision analog signal conversion for each conductive channel in the high-density electromyography sensor array to achieve simultaneous acquisition of all leads; The communication module is used for data communication to achieve data transmission; The main control chip is used to control the operation of the acquisition system; The sensor module is used to obtain attitude information about limb movement; The clock module is used to synchronize the operation of each module to ensure the accuracy of data acquisition; The feedback and display module is used to provide user feedback or display the system status; The driver chip is used to control the feedback and display module; The storage module is used to store the collected electromyogram signal data; The charging and voltage stabilizing module is used to manage the power supply of the acquisition system to ensure the stable operation of each module.

8. A high-density electromyogram acquisition system according to claim 7, characterized in that: A plurality of the analog front-end chips are connected to the main control chip through an SPI interface to form a daisy chain.

9. A high-density electromyogram acquisition system according to claim 7, characterized in that: The communication module includes a wireless communication module and a USB Type-C interface. The wireless communication module is used for data communication to realize wireless data transmission. The USB Type-C interface is used for data communication to realize data transmission, and is reused as the debugging and downloading port of the main control chip, the debugging and downloading port of the wireless communication module, the USB bus of the storage module, and the charging pin of the charging and voltage stabilizing module.

10. A high-density electromyogram acquisition system according to claim 7, characterized in that: The sensor module adopts a 9-axis acceleration sensor; the feedback and display module adopts a vibration motor and an LED display module.

11. A high-density electromyogram acquisition system according to claim 7, characterized in that: The data transmission between the high-density electromyogram acquisition system and the high-density electromyogram sensor array adopts a frame structure, and the frame structure includes a start flag, a sampling sequence number, a plurality of channel data, a test flag bit, and a check bit.

12. A high-density electromyogram gesture recognition method applied to the high-density electromyogram acquisition system according to any one of claims 7 to 11, characterized in that, including the following steps: Collect electromyogram signal data in real time; Perform filtering and denoising and image construction processing on the collected electromyogram signal data; Input the processed electromyogram signal data into an electromyogram gesture recognition model to obtain real-time control decisions; wherein, the network structure of the electromyogram gesture recognition model includes a one-dimensional convolutional neural network, a two-dimensional convolutional neural network, an alternating convolutional neural network, a multi-stream convolutional operation, and a large pooling layer.

13. A high-density electromyogram gesture recognition method according to claim 12, characterized in that: The construction of the electromyogram gesture recognition model includes the following steps: Preprocess the electromyogram signal data; Perform feature extraction, feature selection, and feature dimensionality reduction processing on the preprocessed electromyogram signal data; Input the processed features into the electromyogram gesture recognition model for training; It also includes the following steps: Smooth the prediction results, make multiple judgments within a preset time, and use the multiple votes as control instructions; When the rehabilitation robot is moving, it does not accept control instructions; When the rehabilitation robot is in a closed hand state, it only accepts open hand type instructions.

Images