阅读说明：本技术 基于肌电的佩戴式肌肉阈值监测系统 (Wearable muscle threshold monitoring system based on myoelectricity ) 是由 王洋阳 陈云刚 孟庆典 于 2021-06-07 设计创作，主要内容包括：本发明提出了一种基于肌电的新型佩戴式肌肉阈值监测系统,包括佩戴式肌电采集设备、肢体训练界面、生理状态分析模块和预警。所述的佩戴式肌电采集设备包括有：连接人体的采集部分,用于肌电信号放大和转换的采集模块、用于控制肌电采集以及向生理状态分析模块、传输肌电信号的STM32微处理器、WIFI无线传输电路、以及陀螺仪电路。所述的肢体训练界面需要一部智能手机,用于训练肢体动作。生理状态分析模块通过多种方法识别使用者当前的意图动作,将动作指令发送至预警模块。手机端的预警模块根据接收到的指令进行相应。肌肉信号的采集,分类以及反馈来训练机器监测佩戴者的身体状况,进而预警模块做出相关的响应从而构建成佩戴式肌肉阈值监测系统。(The invention provides a novel wearable muscle threshold monitoring system based on myoelectricity, which comprises wearable myoelectricity acquisition equipment, a limb training interface, a physiological state analysis module and early warning. The wearable myoelectricity acquisition equipment comprises: the system comprises a collecting part connected with a human body, a collecting module used for amplifying and converting myoelectric signals, an STM32 microprocessor used for controlling myoelectric collection and transmitting the myoelectric signals to a physiological state analyzing module, a WIFI wireless transmission circuit and a gyroscope circuit. The limb training interface requires a smart phone for training limb movements. The physiological state analysis module identifies the current intention action of the user through various methods and sends an action instruction to the early warning module. And the early warning module at the mobile phone end corresponds to the received instruction. The acquisition, classification and feedback of muscle signals are used for training a machine to monitor the physical condition of a wearer, and then the early warning module makes a relevant response to construct a wearable muscle threshold monitoring system.)

1. A wearable muscle threshold monitoring system based on myoelectricity comprises: the system comprises wearable myoelectricity acquisition equipment (1), a limb myoelectricity active training interface (2), a physiological state analysis module (3) and an early warning module (4). The system is characterized in that a user collects sEMG electromyographic signals from the hip and the abdomen of a wearer by applying the wearable limb electromyographic acquisition equipment (1) through the limb electromyographic active training interface (2); the physiological state analysis module (3) receives the sEMG electromyographic signals and combines a Recurrent Convolutional Neural Network (RCNN) algorithm to intelligently identify the muscle states of the buttocks and the abdomen of the wearer. In the invention, a mobile phone end software platform for developing data analysis is built, namely, the function of the physiological state analysis module (3) can be realized at a mobile phone end, and the waveform of each channel sEMG electromyography acquisition device can be checked. After the mobile phone terminal utilizes a Recurrent Convolutional Neural Network (RCNN) algorithm to carry out intelligent identification, the early warning device (4) is controlled through the wireless remote device. In addition, the mobile phone end is combined with the network through wireless remote equipment, so that remote monitoring is facilitated.

2. Myoelectric-based wearable muscle threshold monitoring system according to claim 1, wherein the wearable myoelectric acquisition device (1) comprises: the electrode paste and the lead wire thereof (11) are connected in sequence and used for collecting sEMG (electrical stimulation and muscle) electromyographic signals, the bioelectric signal collection module (12) is used for amplifying and converting the electromyographic signals, the STM32 microprocessor (13) and the WIFI wireless data transmission circuit (14) are used for controlling the collection of the electromyographic signals and transmitting the sEMG electromyographic signals to the physiological state analysis module (2), and the system power supply circuit (15) is respectively connected with the bioelectric signal collection module (12) and the STM32 microprocessor (13), wherein the electrode paste in the electrode paste and the lead wire thereof (11) is used for collecting the sEMG electromyographic signals of different muscles of hip and abdomen and is connected with the bioelectric signal collection module (12) through the lead wire and a PJ313B interface and is used for collecting and transmitting the bioelectric signals; the electrode paste is pasted on the hip and the abdomen of a patient, and the muscle measured by the myoelectricity acquisition equipment is as follows: abdominal wall muscles, diaphragm muscles and anal sphincters;

the bioelectrical signal acquisition module (12) is composed of a plurality of bioelectrical signal acquisition chips which are integrated with an analog input AD module with high common mode rejection ratio for receiving muscle voltage acquired by an electrode patch and sEMG (surface-mounted EMG) electromyographic signals, a low-noise programmable gain amplifier for measuring the electromyographic voltage sEMG and amplifying the electromyographic signals and a high-resolution synchronous sampling analog-to-digital converter for converting the analog signals into digital signals;

the STM32 microprocessor (13) is used for adjusting the acquisition mode of the bioelectricity signal acquisition module and adjusting and controlling the WIFI wireless data transmission module (14) to output sEMG (surface acoustic wave) electromyographic signals, and the sEMG electromyographic signals are sent to the mobile phone end to be used for the physiological state analysis module (3) to analyze data and send the analysis results to the early warning equipment (4).

The WIFI wireless data transmission module (14) works in an AP mode, the highest transmission rate is 4Mbps, and under the control of an STM32 microprocessor (13), collected sEMG electromyographic signals are periodically output to the limb electromyographic active training interface (2) and the physiological state analysis module (3) through the WIFI wireless data transmission module (14) in the form of data packets;

the input voltage of the system power supply circuit (16) is 3.7V, the system power supply circuit is powered by a lithium battery (17), and the system power supply circuit provides working voltages of different chips of the system through a voltage conversion module.

According to the myoelectricity-based wearable muscle threshold monitoring system, the early warning device (4) specifically makes different levels of responses at the mobile phone end after receiving the classification result of the physiological state analysis module (3) so as to prevent danger, and in addition, the mobile phone end transmits the results to the acquisition device through the wireless remote device. The acquisition equipment carries out prompts of different levels including voice prompt, vibration prompt and light prompt according to the responses of the different levels.

3. The myoelectric-based wearable muscle threshold monitoring system according to claim 1, wherein the physiological status analysis module (3) has an online real-time classification function capable of classifying the intensity of abdominal and hip muscle movements;

firstly, sEMG electromyographic signals of a tested person are collected through a wearable electromyographic collection device, the collected sEMG electromyographic signals are periodically transmitted to a limb electromyographic active training interface (2) through a WIFI wireless data transmission module (14) and an NRF24L01 wireless remote circuit (17), then an actual voltage value of the sEMG electromyographic signals is analyzed from an A/D conversion result through a conversion algorithm, finally a deep learning parallel Convolutional Neural Network (CNN) model is trained, after training is completed, the sEMG electromyographic signals are transmitted to a physiological state analysis module (3) through the NRF24L01 wireless remote circuit (17), wherein the actual voltage value of the sEMG electromyographic signals is analyzed from the A/D conversion result through the conversion algorithm again, and the actual voltage value is sent to the physiological state analysis module (3) for online real-time classification.

4. The wearable electromyography-based muscle threshold monitoring system of claim 3, wherein the analyzing of the actual voltage value of the sEMG electromyography signal from the a/D conversion result by a conversion algorithm comprises the following steps:

1) determining a reference voltage V of a bioelectrical signal acquisition module (12)_REFAnd the amplification factor G of the programmable gain amplifier_PGA；

2) Converting the original A/D conversion result V of each channel₁₆Converted into decimal A/D conversion result V₁₀；

3) Calculating the actual voltage value V of the sEMG electromyographic signal according to the following formula_IN：

Wherein

5. The myoelectric-based wearable muscle threshold monitoring system according to claim 3, wherein the RCNN algorithm based on the Convolutional Neural Network (CNN) combined with the long-term and short-term memory neural network (LSTM) comprises the following steps:

1) obtaining raw sEMG electromyographic signalsWherein N is the number of channels of the original sEMG electromyographic signals, L is the data length of the original sEMG electromyographic signals of each channel, and E_c,gThe g-th numerical value in the original sEMG electromyographic signals collected by the c-th electrode in the original sEMG electromyographic signals is represented;

2) carrying out digital band-pass filtering on the original sEMG electromyographic signals, carrying out 50Hz notch filtering, removing power frequency interference, and obtaining processed sEMG electromyographic signalsWherein, X_c,gRepresenting the g-th numerical value in the sEMG electromyographic signals corresponding to the c-th electrode after filtering;

3) sEMG electromyographic signal based on digital filteringAfter filtering, a plurality of sample sets are constructed.

4) And entering a parameter fine adjustment stage, and sequentially sending the sEMG electromyographic signals after digital filtering of each user into a convolution neural network model with an initial depth for training and gradient correction.

6. The electromyography-based wearable muscle threshold monitoring system of claim 5, wherein the digital band-pass filtering of the raw sEMG electromyography signals of step 2) is performed by a chebyshev I-type band-pass filter having a first stop band frequency F_stop10.001Hz, first passband frequency F_pass110Hz, second pass band frequency F_pass230Hz, second stop band frequency F_stop2The first stopband attenuation rate is 5dB, and the second stopband attenuation percentage is 5dB at 40 Hz.

7. The wearable electromyography-based muscle threshold monitoring system of claim 5, wherein the digitally filtered sEMG electromyography signal sample sets of each user of step 4) are sequentially fed into an initial depth recursive convolutional neural network model (RCNN). Carrying out full-supervision training on a Recursive Convolutional Neural Network (RCNN), setting the initial learning rate of the model to be 0.004, exponentially attenuating the learning rate to prevent the fixed learning rate from obtaining the optimal model, carrying out 500-period cyclic training, setting the Batchsize to be 128, and setting an early stopping mechanism of Earlystopping. The best model for training the model in all cycles of training is obtained. According to the method, the model parameter fine adjustment is not needed on the basis of pre-training, and when the deep convolutional neural network model is designed, myoelectric data of a plurality of people are collected for training, so that the deep learning model has the cross-tested generalization capability. And the output of the recursive convolutional neural network model is the result which is output by the physiological state analysis module (2) and corresponds to the early warning module (4). Compared with many conventional machine learning algorithms, CNN uses a multi-layer structure to improve generalization and abstraction performance of recognition models, and uses LSTM to perform feature extraction and classification on time series in order to improve tracking accuracy and reduce the number of models. Multiple experiments show that the RCNN algorithm applied in the invention is an efficient method for identifying the motion states of the hip and abdomen muscles in relation to the multi-channel sEMG signal processing;

in the recursive convolutional neural network model, forward propagation and backward propagation are mainly utilized. Firstly, analyzing original sEMG electromyographic signal samples through all CNN layers, and forward propagating sEMG electromyographic signal data sets to obtain output values. Then the error between the output value and the expected value is calculated, and then the error calculation of the new output value and the expected value is carried out through the LSTM. And finally determining the accuracy of the output. Next, the weight values are modified using an error back-propagation process. These two processes are repeatedly performed by the iterative operating system until the loss value of the network is minimized. The weighting values are then modified using a gradient descent algorithm. In order to prevent the final model from being not the optimal model due to excessive modification of the weight, an early stopping mechanism of Earlystopping is arranged in the invention for reducing excessive fitting, and the model with the minimum value of the test loss function is set as the optimal model for training at present.

The recursive convolutional neural network model comprises two parts, wherein in the first part, firstly, digitized and filtered sEMG electromyographic signals are input into a convolutional neural network CNN, the lateral emphasis of the deep convolutional neural network extracts one-dimensional data in a time domain and a frequency domain, the dimensionality of time domain and frequency domain features is reduced in a Maxpooling layer, finally, action prediction results are output, and feature extraction is carried out by utilizing a plurality of convolutional layers and pooling layers. The convolutional layer and the pooling layer of the recurrent convolutional neural network model are deep enough and stable, and can effectively extract the time-frequency domain characteristics of the sEMG electromyographic signals. And then entering a second part, fusing the time domain and frequency domain characteristics which are extracted by the CNN through a long-time memory neural network (LSTM), and extracting the time domain and frequency domain characteristics which are extracted by the CNN in a time sequence mode of internal/external motion. A large number of experiments prove that the method has very good classification effect.

In the recursive convolutional neural network model, the same initialization method is used to initialize the convolution kernel weights and the initial values of the LSTM in the convolutional neural network, so that the output and the input are subjected to the same probability distribution as much as possible. In the recursive convolutional neural network model, L2 regularization is applied to the entire fully-connected layer. A complexity index model is added to the loss function to improve the model's ability to recognize random noise. The model adopts an Adam optimizer, and the loss reduction rule adopts a gradient reduction algorithm.

The Recursive Convolutional Neural Network (RCNN) comprises in order:

(1) the data input layer is used for inputting data which are the sEMG electromyographic signal samples after digital filtering;

(2) the first convolution layer has 32 convolution kernels, the size of which is 1 × 4, and the step size is (1, 1). The convolution kernel selects ReLU as the activation function,

(3) in the second convolutional layer, the number of convolutional kernels is 64, the convolutional kernel size is 4 × 1, and the step size is (1, 1). The convolution kernel selects ReLU as the activation function,

(4) the first maximum pooling layer has the pooling core size of 2 multiplied by 1 and the step length of (1,1), and extracts the maximum value of the elements of the input data covered by the current pooling core as output;

(5) in the third convolutional layer, the number of convolution kernels is 64, the convolution kernel size is 1 × 4, and the step size is (1, 1). The convolution kernel selects ReLU as the activation function,

(6) the second maximum pooling layer, the size of the pooling core is 2 multiplied by 1, the step length is (1,1), and the maximum value of the elements of the input data covered by the current pooling core is extracted and used as output;

(7) the first LSTM layer, with a cell count of 64, selects tanh as the activation function,

(8) the second LSTM layer, with a cell count of 128, selects tanh as the activation function,

(9) the first Dropout layer is used for preventing overfitting from improving the generalization capability of the model, selecting the neuron on the upper layer randomly to enable the neuron not to output, and the selection probability is 0.5;

(10) the node number of the first full-connection layer is 128, the ReLU is selected as an activation function, the L2 norm is selected as a regularization term, and the L2 norm is set to be 0.004;

(11) the second Dropout layer prevents overfitting from improving the generalization capability of the model, randomly selects the neuron on the upper layer to ensure that the neuron does not output, and the selection probability is 0.5;

(12) the second full-link layer outputs the classification result, the number of the selected nodes is 4, Softmax is selected as an activation function, and the Softmax function is a normalized exponential function in nature and is defined asWherein e is a natural logarithm value, z_hFor the output of the h-th neuron, the denominator in the equation acts as a regularization term, such that

Performing feature fusion on the time domain and frequency domain features extracted by the convolutional neural network, extracting time domain and frequency domain features in a brand-new time sequence mode through a long-time and short-time memory neural network (LSTM), entering a full connection layer, and finally classifying;

the wearable electromyography-based muscle threshold monitoring system is characterized in that collected sEMG electromyography signals of 8 channels are taken as a sample 250 times before entering a deep learning algorithm, so that the time domain and frequency domain characteristics of the sEMG electromyography signals are better extracted by the algorithm and sent to the deep learning algorithm;

the wearable electromyography-based muscle threshold monitoring system is characterized in that in the sEMG electromyography signals, electromyography data of each channel can be represented as waveform data of one measured muscle channel. Therefore, in the invention, when the deep learning model is used for semantic segmentation, the conventional spatial convolution pooling of the regular shape NxN is not adopted, but the long and narrow convolution and the strip pooling are adopted. The square convolutional and pooling layers limit the up and down correlation and flexibility for each channel to capture sEMG signals in real-time hand movements.

In order to more effectively capture the correlation and flexibility of the sEMG electromyographic signals to the upper and lower information in each channel, the invention utilizes long and narrow convolution and stripe pooling to expand the receiving range of a deep learning model and collect the correlation characteristics of the upper and lower data information of the sEMG electromyographic signals. In the recurrent convolutional neural network RCNN, according to the frequency characteristics of sEMG electromyographic signals, electromyographic frequencies are mainly distributed in the range of 0-500Hz, and after digital filtering, the frequencies of the sEMG electromyographic signals are all in the frequency range of 0-500Hz, but the frequency distribution characteristics are more remarkable in the muscle states of the buttocks and the abdomen of the collected sEMG electromyographic signals. In the deep learning model, in the first convolution layer, in order to extract the frequency characteristics of the sEMG electromyographic signals, 1 × 4 long and narrow convolution is applied to extract the frequency domain characteristics of the sEMG electromyographic signals. The extracted frequency domain characteristics can effectively represent the multivariate time sequence of the sEMG electromyographic signals, the input sEMG electromyographic signals are subjected to horizontal and vertical long and narrow convolution kernel convolution to calculate H multiplied by 1 and 1 multiplied by W, the H multiplied by 1 and 1 multiplied by W are changed after long and narrow pooling, the maximum value of the element values in the pooling kernel is calculated, and the value is used as a pooling output value. In addition, in the second convolution layer, the time-domain features of the sEMG electromyographic signals are extracted by long and narrow convolution of 4 × 1 size. After the band pooling layer is applied, Maxpooling has local invariance and can extract obvious features and simultaneously reduce parameters of the model, thereby reducing overfitting of the model. The Maxpooling only extracts the significant features of the sEMG electromyographic signals and discards insignificant information, the extracted time-frequency domain features can effectively represent the multivariate time sequence of the sEMG electromyographic signals, and the generation of overfitting can be relieved to a certain extent due to the fact that the parameters of the model are reduced. And sequentially and repeatedly extracting the frequency domain characteristics and the time domain characteristics of the sEMG electromyographic signals in the three convolution layers. And then again do the apply stripe pooling operation. And then enter the long-term memory neural network (LSTM).

The wearable myoelectricity-based muscle threshold monitoring system is characterized in that the calculation formula of the used long and narrow convolution is as follows:

8. use of a myoelectric-based wearable muscle threshold monitoring system of claim 1, a remote drone control module, comprising:

(1) the physiological state analysis module (2) of claim 7 outputs a result, and the result is sent to the early warning module (4) through the WIFI wireless data transmission circuit (14).

9. Use of a myoelectric-based wearable muscle threshold monitoring system according to claim 1, comprising:

(1) in order to ensure that the electrode paste is completely contacted with the skin; before measuring the electromyographic signal sEMG, wiping the skin of the tested position of the dominant hand of the tested person by alcohol cotton, wherein the tested person is required to keep sitting still, annotating a screen on eyes and completing the action required by the limb electromyographic active training interface; after a system power supply circuit (16) is ensured to be normal, the system is started, sEMG electromyographic signals are collected through the wearable electromyographic collection equipment and are transmitted to the physiological state analysis module, and electromyographic collection work is completed;

(2) the physiological state analysis module automatically carries out digital filtering on the collected sEMG electromyographic signals;

(3) the classification and identification system performs classification and identification according to the digitized filtered sEMG electromyographic signals based on a pre-trained recurrent convolutional neural network model;

(4) and sending the classification result to a signal control early warning module (4) through a WIFI wireless data transmission circuit (14).

Technical Field

The invention relates to an intelligent sensing system for muscle motion states. In particular to a wearable muscle threshold monitoring system based on myoelectricity.

Background

The viscera function of the old people is physiologically weakened, the peristalsis capability of the intestinal tract is reduced, and the excrement is easy to be retained in the intestinal tract and cannot be excreted. The rectus and abdominal muscles of the elderly have atrophy and low muscle tension, which causes weakness of defecation and difficult defecation. Therefore, elderly people are prone to constipation, and elderly people over 65 years old are often suffered from chronic diseases such as hypertension and coronary heart disease. When old constipation patients defecate, the abdominal wall muscles and the diaphragm muscles can be strongly contracted due to over-exertion, the abdominal pressure is increased, the blood pressure is further increased, and the heart rate is also accelerated. So the oxygen consumption of the myocardium is increased, the myocardium suffers from serious and persistent acute ischemia, coronary plaque rupture, thrombosis is caused, myocardial infarction is caused, the myocardial infarction is easily caused or blood pressure rises suddenly, cerebral hemorrhage can be caused, the increase of the oxygen consumption of the myocardium can induce angina, myocardial infarction and serious arrhythmia, and sudden death is caused.

In addition, the anal sphincter muscle is relaxed, the tension is reduced and old or some diseases cause atrophy or weakness of contraction in incontinence patients. Therefore, the states of the buttocks and abdominal muscles of the patient with the excrement and urine are monitored, and the defecation pre-prompting of the patient with the old people who exerts too much defecation force and the patient with the incontinence of the excrement and urine is monitored.

With the development of science and technology, the intelligent identification and application of electromyographic signals gradually enter daily life, and the intelligent identification and application of electromyographic signals become an embodiment of intelligent life. The human muscle electric signal is gathered, and categorised discernment is carried out, understands human muscle intention, and then makes relevant response, becomes an important trend of intelligent life gradually.

Whenever cells in the hand are stimulated by an electric shock or are activated by nerves, the body generates corresponding muscle potentials. Therefore, the electrical signals of the muscles of the human body are measured at this time, so that the corresponding muscle movement of the human body can be detected and analyzed, or the level of the stimulated neurons can be known. Therefore, when the muscle contracts or expands, the transmitted electric signals can reflect the movement conditions of the nerve and the muscle to a certain extent, and then the training classification is carried out by utilizing an algorithm. With the rise of deep learning and the improvement of computing power, the real-time judgment of human body movement becomes possible according to the classification judgment of the algorithm.

The technology of collecting human body electric signals is a new way to establish information channel between human body and computer. The human body electric signal acquisition technology is used for acquiring and analyzing hand electromyographic signals of a testee, extracting rich characteristics contained in the electromyographic signals, further judging the hand action state of the testee, and possibly being used for prosthesis motion control, clinical hand disease detection, clinical diagnosis of motion injury, medical detection and medical rehabilitation and improvement of daily life activities in the future, even in some game and entertainment fields.

However, most of the existing multi-channel myoelectricity acquisition devices have the inconveniences of high price, heavy volume, complex operation, incapability of real-time classification and the like, and the wearable myoelectricity acquisition devices generally have the defects of insufficient precision, fewer channels and the like. Therefore, it is necessary to design and develop a set of high-precision and multi-channel wearable electromyography acquisition equipment and use the equipment in the field of human body electric signal acquisition.

An embedded single chip microcomputer (STM32) microprocessor with high performance, low cost and low power consumption is widely applied to the application fields of industrial control, consumer electronics, Internet of things, communication equipment, medical service, security monitoring and the like. Wearing formula flesh electricity collection equipment based on STM32 stability and the high performance theory design of operation has solved the passageway less, wireless transmission scheduling problem. Therefore, the STM32 is used as a main control chip of the wearable myoelectricity acquisition equipment and matched with the high-precision bioelectricity signal acquisition module, and the acquisition precision of the equipment can be met.

Electromyographic signals are a bioelectrical signal that is generated from any tissue or organ, typically a function of time and a series of amplitudes, frequencies, and waveforms. The electromyographic signals are bioelectric signals generated along with muscle contraction, and the collected electromyographic signals on the skin surface are called surface electromyographic signals sEMG. sEMG electromyographic signals are bioelectric currents generated by contraction of the muscles of the surface of the human body. The nervous system controls the activity of the muscles (contraction or relaxation) and the different muscle fiber motor units in the surface skin generate mutually different signals at the same time. Therefore, the electromyographic signals have the characteristics of nonlinearity, unsteadiness, serious noise interference and the like.

In recent years, deep learning has shown its powerful potential in the fields of object detection, speech recognition, and natural language processing. The deep convolutional neural network is a representative of successful applications in deep learning, and can effectively extract features in the grid-like data. The models of the self-coding neural network and the deep convolution neural network of the multi-layer neurons can exert the corresponding advantages of the respective components, extract the time domain and frequency domain characteristics of the corresponding electromyographic signals, and further realize accurate identification of hand movements of a testee.

Disclosure of Invention

The invention aims to solve the technical problem of providing a wearable muscle threshold monitoring system based on myoelectricity, which can rapidly and accurately perform primary identification on the states of buttocks and abdominal muscles of a patient so as to prompt early warning to protect the safety of a wearer.

The technical scheme adopted by the invention is as follows: wearable equipment based on ADS of low-power consumption gathers chip design is used for gathering human body surface muscle electric signal (sEMG), and in the acquisition process, sEMG signal is through filtering, enlargies, and the digitization is transmitted to the cell-phone end through wireless transmission equipment again, in the dedicated APP of cell-phone, carries out digital filtering, and the application deep learning algorithm trains and classifies. This research has set up the cell-phone end software platform that has developed data analysis, can be fast, accurate classify to the state of patient's buttock and abdomen muscle, and classification effect can reach higher level to in the APP of cell-phone end, can look over the real-time waveform of sEMG flesh signal. After the signal is collected, the predicted signal can be transmitted to the mobile phone end for further monitoring. In addition, the monitoring can be carried out at the mobile phone end, so that the real-time dynamics of the wearer can be conveniently known. The higher accuracy rate indicates that the set of system functions is feasible. In the invention, a detailed algorithm structure based on neural network training is given.

Drawings

FIG. 1 is a block diagram of the wearable electromyography-based muscle threshold monitoring system of the present invention;

FIG. 2 is a block diagram of the wearable electromyography acquisition device of the present invention;

FIG. 3 is a block diagram of a wireless transmission module according to the present invention;

FIG. 4 is a block diagram of the RCNN algorithm of the present invention;

Detailed Description

The myoelectric-based wearable muscle threshold monitoring system of the present invention will be described in detail below with reference to examples and the accompanying drawings.

As shown in fig. 1, the wearable myoelectric-based muscle threshold monitoring system includes: the system comprises wearable myoelectricity acquisition equipment (1), a limb myoelectricity active training interface (2), a physiological state analysis module (3) and an early warning module (4). The hand myoelectricity active training device is characterized in that a user collects sEMG (surface EMG) myoelectricity signals from the hand of the user by applying the wearable hand myoelectricity collecting device (1) through the hand myoelectricity active training interface (2).

The physiological state analysis module (3) receives the sEMG electromyographic signals and combines a Recurrent Convolutional Neural Network (RCNN) algorithm to intelligently identify the muscle states of the buttocks and the abdomen of the wearer. In the invention, a mobile phone end software platform for developing data analysis is built, namely, the function of the physiological state analysis module (3) can be realized at a mobile phone end, and the waveform of each channel sEMG electromyography acquisition device can be checked. After the mobile phone terminal utilizes a Recurrent Convolutional Neural Network (RCNN) algorithm to carry out intelligent identification, the early warning device (4) is controlled through the wireless remote device. In addition, the mobile phone end is combined with the network through wireless remote equipment, so that the remote monitoring of family members is facilitated.

The early warning module (4) takes the physiological state analysis module (3) as a basis, and controls the early warning device (4) to make a response corresponding to the muscle state of the limb through wireless remote equipment. The prompt of the early warning device (4) comprises a voice prompt, a vibration prompt and a light prompt.

As shown in fig. 2, the wearable electromyography-based muscle threshold monitoring system is characterized in that the wearable electromyography acquisition device (1) comprises: the electrode paste and the lead wire thereof (11) are connected in sequence and used for collecting sEMG electromyographic signals, a bioelectric signal collection module (12) used for amplifying and converting the electromyographic signals, an STM32 microprocessor (13) and a WIFI wireless data transmission circuit (14) used for controlling the collection of the electromyographic signals and transmitting the sEMG electromyographic signals to the physiological state analysis module (2), and a system power supply circuit (15) respectively connected with the bioelectric signal collection module (12) and the STM32 microprocessor (13), wherein the electrode paste in the electrode paste and the lead wire thereof (11) is used for collecting sEMG electromyographic signals of different muscles of hip and abdomen and is connected with the bioelectric signal collection module (12) through the lead wire and a PJ313B interface and used for collecting and transmitting the bioelectric signals; the electrode paste is pasted on the hip and the abdomen of a patient, and the muscle measured by the myoelectricity acquisition equipment is as follows: abdominal wall muscles, diaphragm muscles and anal sphincters.

The bioelectrical signal acquisition module (12) is composed of a plurality of bioelectrical signal acquisition chips which are integrated with a high common mode rejection ratio analog input AD module for receiving human body surface muscle voltage sEMG signals acquired by the electrode patches, a low-noise programmable gain amplifier for measuring the myoelectric voltage sEMG and amplifying myoelectric signals and a high-resolution synchronous sampling analog-digital converter for converting the analog signals into digital signals.

The STM32 microprocessor (13) is used for adjusting the acquisition mode of the bioelectricity signal acquisition module and adjusting and controlling the WIFI wireless data transmission module (14) to output sEMG (surface acoustic wave) electromyographic signals, and the sEMG electromyographic signals are sent to the mobile phone end to be used for the physiological state analysis module (3) to analyze data and send the analysis results to the early warning equipment (4).

The WIFI wireless data transmission module (14) works in an AP mode, the highest transmission rate is 4Mbps, and under the control of an STM32 microprocessor (13), collected sEMG electromyographic signals are periodically output to a limb electromyographic active training interface (2) and a physiological state analysis module (3) at a mobile phone end in the form of data packets through the WIFI wireless data transmission module (14);

the input voltage of the system power supply circuit (15) is 3.7V, the lithium battery (16) supplies power, and the working voltages of different chips of the system provided by the voltage conversion module are-2.5V, 2.5V and 3.3V respectively.

According to the myoelectricity-based wearable muscle threshold monitoring system, the early warning device (4) specifically makes different levels of responses at the mobile phone end after receiving the classification result of the physiological state analysis module (3) so as to prevent danger, and in addition, the mobile phone end transmits the results to the acquisition device through the wireless remote device. The acquisition equipment carries out prompt of different levels including voice prompt, vibration prompt and light prompt according to the response of different levels.

The wearable myoelectricity-based muscle threshold monitoring system is characterized in that the physiological state analysis module (3) has an online real-time classification function, and the real-time online classification function can classify the intensity of muscle movement of the abdomen and the hip.

Firstly, sEMG electromyographic signals of a tested person are collected through a wearable electromyographic collecting device, the collected sEMG electromyographic signals are periodically transmitted to a limb electromyographic active training interface (2) through a WIFI wireless data transmission module (14), then actual voltage values of the sEMG electromyographic signals are analyzed from A/D conversion results through a conversion algorithm, finally a deep learning Convolutional Neural Network (CNN) model is trained, after training is completed, the sEMG electromyographic signals are transmitted to a physiological state analysis module (3) through the WIFI wireless data transmission module (14), wherein the actual voltage values of the sEMG electromyographic signals are analyzed from the A/D conversion results through the conversion algorithm again, and the actual voltage values are transmitted to the physiological state analysis module (3) for online real-time classification.

The wearable electromyography-based muscle threshold monitoring system is characterized in that the actual voltage value of the sEMG electromyography signal is analyzed from the A/D conversion result through a conversion algorithm, and the system comprises the following steps:

1) determining a reference voltage V of a bioelectrical signal acquisition module (12)_REFAnd the amplification factor G of the programmable gain amplifier_PGA；

2) Converting the original A/D conversion result V of each channel₁₆Converted into decimal A/D conversion result V₁₀；

3) According to the following formula, sEMG electromyographic signals are calculatedActual voltage value V_IN：

Wherein

The wearable myoelectricity-based muscle threshold monitoring system is characterized in that the RCNN algorithm of the recurrent neural network based on the combination of the recurrent neural network (CNN) and the long-time memory neural network (LSTM) comprises the following steps:

1) obtaining raw sEMG electromyographic signalsThe method comprises the steps that N is the number of channels of original sEMG electromyographic signals, L is the data length of the original sEMG electromyographic signals of each channel, and represents the g-th numerical value of the original sEMG electromyographic signals collected by the c-th electrode in the original sEMG electromyographic signals;

2) carrying out digital band-pass filtering on the original sEMG electromyographic signals, carrying out 50Hz notch filtering, removing power frequency interference, and obtaining processed sEMG electromyographic signalsWherein, X_c,gRepresenting the g-th numerical value in the sEMG electromyographic signal corresponding to the c-th electrode after filtering;

3) sEMG electromyographic signal based on digital filteringAfter filtering, a plurality of sample sets are constructed.

4) And entering a parameter fine-tuning stage, sequentially sending the sEMG electromyographic signals after digital filtering of each user into a Recursive Convolutional Neural Network (RCNN) model with initial depth, and performing training and gradient correction.

The invention relates to a wearable myoelectricity-based muscle threshold monitoring system, which is characterized in that,the step 2) of digitally filtering the original sEMG electromyographic signals adopts a band-pass filter, wherein a first stop band frequency F of the band-pass filter_stop10.001Hz, first passband frequency F_pass110Hz, second pass band frequency F_pass230Hz, second stop band frequency F_stop2The first stopband attenuation rate is 5dB, and the second stopband attenuation percentage is 5dB at 40 Hz.

The wearable electromyography-based muscle threshold monitoring system is characterized in that each user digitally filtered sEMG electromyography signal sample set in the step 4) is sequentially sent to a Recursive Convolutional Neural Network (RCNN) model at an initial depth. Carrying out full-supervision training on a Recursive Convolutional Neural Network (RCNN), setting the initial learning rate of the model to be 0.004, exponentially attenuating the learning rate to prevent the fixed learning rate from obtaining the optimal model, carrying out 500-period cyclic training, setting the Batchsize to be 128, and setting an early stopping mechanism of Earlystopping. The best model for training the model in all cycles of training is obtained. According to the invention, when a recurrent convolutional neural network model is designed, myoelectric data of a plurality of people are collected for training, so that the deep learning model has the generalization capability across the tested object. And the output of the recursive convolutional neural network model is the result which is output by the physiological state analysis module (2) and corresponds to the early warning module (4). Compared with many conventional machine learning algorithms, CNN uses a multi-layer structure to improve generalization and abstraction performance of recognition models, and uses LSTM to perform feature extraction and classification on time series in order to improve tracking accuracy and reduce the number of models. Multiple experiments show that the RCNN algorithm applied in the invention is an efficient method for identifying the motion state of the hip and abdomen muscles in relation to multi-channel sEMG signal processing.

In the recursive convolutional neural network model, forward propagation and backward propagation are mainly utilized. Firstly, analyzing original sEMG electromyographic signal samples through all CNN layers, and forward propagating sEMG electromyographic signal data sets to obtain output values. Then the error between the output value and the expected value is calculated, and then the error calculation of the new output value and the expected value is carried out through the LSTM. And finally determining the accuracy of the output. Next, the weight values are modified using an error back-propagation process. These two processes are repeatedly performed by the iterative operating system until the loss value of the network is minimized. The weighting values are then modified using a gradient descent algorithm. In order to prevent the final model from being not the optimal model due to excessive modification of the weight, an early stopping mechanism of Earlystopping is arranged in the invention for reducing excessive fitting, and the model with the minimum value of the test loss function is set as the optimal model for training at present.

The recursive convolutional neural network model comprises two parts, wherein in the first part, firstly, digitized and filtered sEMG electromyographic signals are input into a convolutional neural network CNN, the lateral emphasis of the deep convolutional neural network extracts one-dimensional data in a time domain and a frequency domain, the dimensionality of time domain and frequency domain features is reduced in a Maxpooling layer, finally, action prediction results are output, and feature extraction is carried out by utilizing a plurality of convolutional layers and pooling layers. The convolutional layer and the pooling layer of the recurrent convolutional neural network model are deep enough and stable, and can effectively extract the time-frequency domain characteristics of the sEMG electromyographic signals. And then entering a second part, fusing the time domain and frequency domain characteristics which are extracted by the CNN through a long-time and short-time memory neural network (LSTM), and extracting the time domain and frequency domain characteristics which are extracted by the CNN in a time sequence mode of internal/external motion. A large number of experiments prove that the method has very good classification effect.

In the recursive convolutional neural network model, the same initialization method is used to initialize the convolutional kernel weights and the initial values of the LSTM in the convolutional neural network, so that the output and the input are subjected to the same probability distribution as much as possible. In the recursive convolutional neural network model, L2 regularization is applied to the entire fully-connected layer. A complexity index model is added to the loss function to improve the ability of the model to recognize random noise. The model adopts an Adam optimizer, and the loss reduction rule adopts a gradient reduction algorithm.

The Recursive Convolutional Neural Network (RCNN) comprises in order:

(1) the data input layer is used for inputting data which are the sEMG electromyographic signal samples after digital filtering;

(2) the first convolution layer has 32 convolution kernels, the size of which is 1 × 4, and the step size is (1, 1). The convolution kernel selects ReLU as the activation function,

(3) in the second convolutional layer, the number of convolutional kernels is 64, the convolutional kernel size is 4 × 1, and the step size is (1, 1). The convolution kernel selects ReLU as the activation function,

(4) the first maximum pooling layer has the pooling core size of 2 multiplied by 1 and the step length of (1,1), and extracts the maximum value of the elements of the input data covered by the current pooling core as output;

(5) in the third convolutional layer, the number of convolution kernels is 64, the convolution kernel size is 1 × 4, and the step size is (1, 1). The convolution kernel selects ReLU as the activation function,

(6) the second maximum pooling layer, the size of the pooling core is 2 multiplied by 1, the step length is (1,1), and the maximum value of the elements of the input data covered by the current pooling core is extracted and used as output;

(7) the first LSTM layer, with a cell count of 64, selects tanh as the activation function,

(8) the second LSTM layer, with a cell count of 128, selects tanh as the activation function,

(9) the first Dropout layer is used for preventing overfitting from improving the generalization capability of the model, selecting the neuron on the upper layer randomly to enable the neuron not to output, and the selection probability is 0.5;

(10) the first full-connection layer selects the number of nodes as 128, selects ReLU as an activation function, selects an L2 norm as a normalizing term, and sets an L2 norm as 0.004;

(11) the second Dropout layer prevents overfitting from improving the generalization capability of the model, randomly selects the neuron on the upper layer to ensure that the neuron does not output, and the selection probability is 0.5;

(12) the second full-link layer outputs the classification result, the number of the selected nodes is 4, Softmax is selected as an activation function, and the Softmax function is a normalized exponential function in nature and is defined asWherein e is a natural logarithm value, z_hFor the output of the h-th neuron, the denominator in the equation acts as a regularization term, such thatAnd (3) performing feature fusion on the time domain and frequency domain features extracted by the convolutional neural network, extracting time domain and frequency domain features in a brand-new time sequence mode through a long-time and short-time memory neural network (LSTM), entering a full connection layer, and finally classifying. The output of the full connection layer is the output of the physiological state analysis module (2), as shown in fig. 4.

The wearable electromyography-based muscle threshold monitoring system is characterized in that collected sEMG electromyography signals of 8 channels are taken as a sample 250 times before entering a deep learning algorithm, so that the time domain and frequency domain characteristics of the sEMG electromyography signals are better extracted by the algorithm and are sent to the deep learning algorithm.

The wearable electromyography-based muscle threshold monitoring system is characterized in that in the sEMG electromyography signals, electromyography data of each channel can be represented as waveform data of one measured muscle channel. Therefore, in the invention, when the deep learning model is used for semantic segmentation, the conventional spatial convolution pooling of the regular shape NxN is not adopted, and the narrow convolution and the strip pooling are adopted. The square convolutional and pooling layers limit the up and down correlation and flexibility for each channel to capture sEMG signals in real-time hand movements.

In order to more effectively capture the correlation and flexibility of the sEMG electromyographic signals to the upper and lower information in each channel, the receiving range of a deep learning model is expanded by utilizing long and narrow convolution and banding pooling, and the correlation characteristics of the upper and lower information of the sEMG electromyographic signals are collected. In the recurrent convolutional neural network RCNN, myoelectric frequencies are mainly distributed in the range of 0-500Hz according to the frequency characteristics of sEMG electromyographic signals, and after digital filtering, the frequencies of the sEMG electromyographic signals are all in the frequency range of 0-500Hz, but the frequency distribution characteristics are more remarkable in the muscle states of the buttocks and the abdomen of the collected sEMG electromyographic signals. In the deep learning model, in the first convolution layer, in order to extract the frequency characteristics of the sEMG electromyographic signals, 1 × 4 long and narrow convolution is applied to extract the frequency domain characteristics of the sEMG electromyographic signals. The extracted frequency domain characteristics can effectively represent the multivariate time sequence of the sEMG electromyographic signals, the inputted sEMG electromyographic signals are subjected to horizontal and vertical long and narrow convolution kernel convolution to calculate H multiplied by 1 and 1 multiplied by W, the H multiplied by 1 and 1 multiplied by W are changed after long and narrow pooling, the maximum value of the element values in the pooling kernel is obtained, and the value is used as a pooling output value. In addition, in the second convolution layer, the time-domain features of the sEMG electromyographic signals are extracted by long and narrow convolution of 4 × 1 size. After the band pooling layer is applied, Maxpooling has local invariance and can extract obvious features and simultaneously reduce the parameters of the model, thereby reducing overfitting of the model. The Maxpooling extracts only the significant features of the sEMG electromyographic signals and discards insignificant information, the extracted time-frequency domain features can effectively represent the multivariate time sequence of the sEMG electromyographic signals, and the generation of overfitting can be relieved to a certain extent due to the fact that the parameters of the model are reduced. And sequentially and repeatedly extracting the frequency domain characteristics and the time domain characteristics of the sEMG electromyographic signals in the three convolution layers. And then again do the apply stripe pooling operation. And then enter a long-term and short-term memory neural network (LSTM).

The wearable myoelectricity-based muscle threshold monitoring system is characterized in that the calculation formula of the used long and narrow convolution is as follows:

the invention relates to an application of a wearable myoelectricity-based muscle threshold monitoring system, a pre-warning module (4), which is characterized by comprising:

(1) the physiological state analysis module (2) of claim 7 outputs a result, and the result is sent to the early warning module (4) through the WIFI wireless data transmission circuit (14).

The invention discloses application of a wearable myoelectricity-based muscle threshold monitoring system, which is characterized by comprising the following components:

(1) in order to ensure that the electrode paste is completely contacted with the skin; before measuring the electromyographic signal sEMG, wiping the skin of the tested position of the dominant hand of the tested person by alcohol cotton, wherein the tested person is required to keep sitting still, annotating a screen on eyes and completing the action required by the limb electromyographic active training interface; after the system power supply circuit (16) is ensured to be normal, the system is started, sEMG electromyographic signals are collected through the wearable electromyographic collection equipment and are transmitted to the physiological state analysis module, and electromyographic collection work is completed;

(2) the physiological state analysis module automatically carries out digital filtering on the collected sEMG electromyographic signals;

(3) the classification and identification system performs classification and identification according to the digitized filtered sEMG electromyographic signals based on a pre-trained recurrent convolutional neural network model;

(4) and sending the classification result to a signal control early warning module (4) through a WIFI wireless data transmission circuit (14).