WO2020118797A1

WO2020118797A1 - Prosthesis control method, apparatus, system and device, and storage medium

Info

Publication number: WO2020118797A1
Application number: PCT/CN2018/125437
Authority: WO
Inventors: 黄品高; 李光林; 张元康; 翁恭伟; 黄天展; 杨子健; 王辉; 于文龙; 黄剑平; 汪圆圆
Original assignee: 深圳先进技术研究院
Priority date: 2018-12-13
Filing date: 2018-12-29
Publication date: 2020-06-18
Also published as: CN111317600A; CN111317600B

Abstract

Provided are a prosthesis control method, apparatus, system and device, and a storage medium. The method comprises: acquiring a muscle signal at a prosthetic joint, wherein the muscle signal comprises an electromyographic signal and a muscle impedance signal (110); recognizing an action type corresponding to the muscle signal (120); and generating a prosthesis control instruction according to the action type (130).

Description

Prosthetic limb control method, device, system, equipment and storage medium

This disclosure requires the priority of a Chinese patent application filed with the Chinese Patent Office on December 13, 2018, with application number 201811524294.2, the entire contents of which are incorporated by reference in this disclosure.

Technical field

The present disclosure relates to physiological signal detection technology, for example, to a prosthetic limb control method, device, system, equipment, and storage medium.

Background technique

With the continuous development of EMG signal detection and recognition technology, EMG signals have been widely researched and applied in prosthetic limb control, muscle disease diagnosis, and nervous system disease analysis. EMG signals are physiological electrical signals transmitted by a large number of motor units excited by electrical sequences of movement along muscle fibers. Different types of limb movements have different muscle contraction patterns. The differences in these patterns are reflected in the differences in the characteristics of EMG signals. These differences are used to distinguish different types of muscle movements, thus making the prosthetic movements more natural and easier to control. However, because the EMG signal is relatively weak and susceptible to interference, it is difficult to collect and process the EMG signal in real time

Therefore, the accuracy rate of prosthetic limb control based on EMG signals is low.

Summary of the invention

The present disclosure provides a prosthetic limb control method, device, system, equipment and storage medium to improve the accuracy of prosthetic limb control.

An embodiment of the present disclosure provides a prosthetic limb control method. The method includes:

Obtain the muscle signal at the joint of the prosthesis, where the muscle signal includes myoelectric signal and muscle impedance signal;

Identify the action type corresponding to the muscle signal;

The control instruction of the prosthesis is generated according to the action type.

In an embodiment, the action type corresponding to the recognition muscle signal includes:

Superimposing the myoelectric signal and the muscle impedance signal to obtain a superimposed muscle signal;

The superimposed muscle signal is input into a pre-trained muscle action type recognition model to obtain an action type corresponding to the muscle signal.

Extracting the characteristics of the EMG signal to obtain the EMG characteristic signal, and extracting the characteristics of the muscle impedance signal to obtain the muscle impedance characteristic signal;

Superimposing the EMG characteristic signal and the muscle impedance characteristic signal to obtain a superimposed characteristic signal;

The superimposed feature signal is input into a pre-trained feature action type recognition model to obtain the action type corresponding to the muscle signal.

Input the EMG signal into a pre-trained EMG action type recognition model to obtain the EMG action type;

Input the muscle impedance signal into a pre-trained muscle impedance action type recognition model to obtain the muscle impedance action type;

The action type corresponding to the muscle signal is determined according to the myoelectric action type and the muscle impedance action type.

In an embodiment, the muscle action type recognition model is trained in the following manner:

Obtain the EMG sample signal, muscle impedance sample signal and the action type corresponding to the EMG sample signal and muscle impedance signal;

Superimposing the EMG sample signal and the muscle impedance sample signal to obtain a superimposed muscle sample signal;

Taking the superimposed muscle sample signal as an input variable, the action type corresponding to the myoelectric sample signal and the muscle impedance sample signal as output variables, training a classifier model, and obtaining the muscle action type recognition model.

In an embodiment, the characteristic action type recognition model is trained in the following manner:

Extracting the characteristics of the EMG sample signal to obtain the EMG sample characteristic signal, and extracting the characteristics of the muscle impedance sample signal to obtain the muscle impedance sample characteristic signal;

Superimposing the EMG sample characteristic signal and the muscle impedance sample characteristic signal to obtain a superimposed muscle sample characteristic signal;

Using the characteristic signal of the superimposed muscle sample as an input variable, the action type corresponding to the EMG sample signal and the muscle impedance sample signal as the output variable, and training the classifier model to obtain the training that the action type corresponding to the muscle signal is the output variable Classifier model, get the characteristic action type recognition model.

In one embodiment, the EMG type recognition model is trained in the following manner:

Obtain the EMG sample signal, the muscle impedance sample signal, and the action type corresponding to the EMG sample signal and the muscle impedance sample signal;

Obtain the EMG sample signal and the action type corresponding to the EMG sample signal;

Using the EMG sample signal as an input variable and the action type corresponding to the EMG sample signal as an output variable, training a classifier model to obtain the EMG action type recognition model;

Train the muscle impedance action type recognition model by:

Obtain the muscle impedance sample signal and the action type corresponding to the muscle impedance sample signal;

Using the muscle impedance sample signal as an input variable and the action type corresponding to the muscle impedance sample signal as an output variable, a classifier model is trained to obtain the muscle impedance action type recognition model.

In one embodiment, the EMG signal is obtained after processing according to the muscle impedance signal.

In an embodiment, the EMG signal is obtained after processing according to the muscle impedance signal, including:

Obtain the EMG signal to be screened, the EMG signal to be screened and the muscle impedance signal are acquired through synchronization;

If the muscle impedance signal is less than a preset threshold, the EMG signal to be screened is used as the EMG signal.

An embodiment of the present disclosure also provides a prosthetic limb control device, which includes:

A muscle signal acquiring module, configured to acquire a muscle signal at the connection of the prosthesis, wherein the muscle signal includes an electromyographic signal and a muscle impedance signal;

An action type recognition module configured to recognize the action type corresponding to the muscle signal;

The control instruction generation module is configured to generate the control instruction of the prosthesis according to the action type.

An embodiment of the present disclosure also provides a prosthetic limb control system, which includes an upper computer and a lower computer. The upper computer is provided with a prosthetic limb control device as described in an embodiment of the present disclosure; the upper computer communicates with the lower computer ;

The lower computer is configured to collect muscle signals, the muscle signals include myoelectric signals and muscle impedance signals, and send the muscle signals to the upper computer.

In an embodiment, the lower computer includes an acquisition module and a control module, and the control module is connected to the acquisition module and the upper computer respectively;

The control module is configured to control the collection module to synchronously collect the muscle signal and the muscle impedance signal, and send the myoelectric signal and the muscle impedance signal to the host computer.

In an embodiment, the acquisition module includes an electrode unit and an analog front-end unit, the analog front-end unit includes a modulator, a first channel, and a second channel; the electrode unit is configured to be placed on the surface of the muscle tissue where the prosthesis is connected , The electrode unit is respectively connected to the modulator, the first channel and the second channel;

The modulator is configured to send the modulated signal to the electrode unit so that the electrode unit collects a mixed signal including the myoelectric signal and the muscle impedance signal, and the first channel is set to The mixed signal is processed to obtain the myoelectric signal, and the second channel is configured to process the mixed signal to obtain the muscle impedance signal.

An embodiment of the present disclosure also provides a device, which includes:

One or more processors;

Memory, set to store one or more programs;

When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in the embodiments of the present disclosure.

An embodiment of the present disclosure also provides a computer-readable storage medium that stores a computer program, and when the program is executed by a processor, the method described in the embodiment of the present disclosure is implemented.

The technical solution provided by the present disclosure improves the accuracy of prosthetic limb control by acquiring muscle signals at the joint of the prosthetic limb, including the electromyographic signal and the muscle impedance signal, identifying the action type corresponding to the muscle signal, and generating control commands for the prosthetic limb according to the action type .

BRIEF DESCRIPTION

FIG. 1 is a flowchart of a prosthetic limb control method in an embodiment;

2 is a schematic structural diagram of an artificial limb control device in an embodiment;

3 is a schematic structural diagram of a prosthetic limb control system in an embodiment;

4 is a schematic structural diagram of an operation interface of a main window of a host computer in an embodiment;

5 is a schematic structural diagram of an analog front-end unit in an embodiment;

FIG. 6 is a schematic structural diagram of a device in an embodiment.

detailed description

The disclosure will be described below with reference to the drawings and embodiments. The specific embodiments described herein are only used to explain the present disclosure, rather than to limit the present disclosure. In addition, for convenience of description, only some parts but not all structures related to the present disclosure are shown in the drawings.

Because the EMG signal is weak and susceptible to interference, in order to obtain better results, the EMG signal is often mixed with other signals for joint processing and analysis. Muscle impedance is a comprehensive impedance composed of resistance and capacitance in the blood, muscle and cell tissue of the organism. The muscle impedance applies an excitation current to the region of the muscle tissue under test through the electrode, detects the voltage signal of the corresponding region and is calculated by Ohm's law. As an important parameter of muscle characteristics, it is also commonly used for muscle disease analysis. EMG signals can be combined with muscle impedance signals for joint processing and analysis.

Since most acquisition devices can only realize single signal acquisition of EMG signals and muscle impedance signals, it is difficult to perform combined processing and analysis of EMG signals combined with muscle impedance signals.

Example one

FIG. 1 is a flowchart of a prosthetic limb control method provided by an embodiment. This embodiment can be applied to a case where a prosthetic limb is performed based on myoelectric signals and muscle impedance signals. The method can be performed by a prosthetic limb control device, which can Implemented in software and/or hardware, the device can be configured in a device, such as a computer. As shown in Figure 1, the method includes the following steps.

Step 110: Obtain the muscle signal at the connection of the prosthesis. The muscle signal includes the myoelectric signal and the muscle impedance signal.

In the embodiment, the prosthetic limb is a bionic limb that can be combined with microelectronic technology, computer control technology, biomedical engineering technology, sensor technology, etc., and can imitate the feeling and movement of a human limb. The ultimate goal of prosthetic limb research is to create a humanoid limb with a shape similar to that of a human limb, a function close to that of a human limb, a feeling similar to that of a human limb, and a real-time control of at least one movement. In the human body, muscles, as an important part of the sports system, can convert chemical energy into mechanical energy, and various movements of the human body are more or less related to muscles.

Muscle signals can include myoelectric signals and muscle impedance signals, where the electromyographic signals record changes in the potential of skeletal muscles under different functional states. This potential change is related to muscle structure, contraction strength, and chemical changes during contraction. The mechanism of myoelectric signal generation is that the myoelectric signal originates from the motor nerve unit in the spinal cord as a part of the central nerve. The cell body of the motor nerve unit is in the spinal cord. The axon of the cell body extends to the muscle fiber and passes through the endplate area. Muscle fiber coupling (the coupling of the nature of biochemical processes). There is more than one muscle fiber associated with each neuron, and these parts combine to form a so-called motor unit. The potential difference between the inside and outside of the muscle cell membrane is called the transmembrane potential or membrane potential. When the muscle is quiet, the phenomenon is negative inside the membrane and positive outside the membrane is called polarization. The potential difference between the inside and outside of the muscle cell membrane is called the resting potential. When muscle cells are excited, the membrane potential changes in depolarization and repolarization, and expands to the surrounding, called the action potential. The action potential of a single muscle fiber is called a single fiber action potential. Multiple muscle fibers controlled by the same nerve branch and the nerve branch together form a motor unit. When muscles are stimulated from the central nerve or externally, all muscle fibers of the same motor unit are excited synchronously. The action potentials of all muscle cells of the same motor unit are collectively called exercise. Unit action potential. There are four different types of biopotentials in skeletal muscle cells: resting potential, action potential, terminal potential and injury potential. Under the control of the central nervous system, motor neurons generate electrical pulses, which are transmitted to muscle fibers along the axon, and cause a pulse sequence on all muscle fibers to propagate in two directions along the muscle fibers. These electrical pulses cause the muscle fibers to contract and produce muscle tension. At the same time, the propagating electrical pulses cause a current field in the human soft tissue and show a potential difference between the detection electrodes. The polarity of the potential waveform displayed by multiple muscle fibers at the detection point is related to the relative position of the end plate and the detection point, and to the distance between the fiber and the measurement point: the farther the distance, the smaller the amplitude. The sum of potentials caused by multiple muscle fibers between the detection points constitutes the action potential of the motor unit. Since the electrical distribution on the axon is a pulse sequence, the action potential sequence is also generated between the detection points, and the EMG signal is the sum of the action potential sequence generated by many motor units. According to the type and placement of the detection electrodes, EMG signals can be divided into two types: needle electrode EMG signal and surface electrode EMG signal (ie surface EMG signal). The former uses needle electrode as the guide electrode. Implanted into the muscle, the electrical activity detected directly near the active muscle fiber, the latter uses the surface electrode as the guide electrode, and the electrical activity of the muscle picked up when the surface electrode is placed on the skin surface is synthesized at the detection surface. The surface electrode has the advantage of non-invasive measurement. According to the needs and actual conditions, any number of surface EMG signals can be collected at any location on the skin surface at any time without the participation of a doctor or professional nursing staff, and the duration of the collected signal is within the tolerance range The internal can be controlled freely. Therefore, the EMG signal described in this embodiment refers to the surface EMG signal detected by the surface electrode. Different movements of the limb correspond to different muscle contraction patterns. The differences in these patterns can be reflected in the difference in surface EMG signals. Therefore, the surface EMG signals can be analyzed to distinguish the different movement types of the limbs in order to control the prosthesis. . The principle of prosthetic limb control based on surface myoelectric signals is as follows: the information source of surface prosthetic signals to control prosthetic limbs is the remaining muscle groups of the residual limb. Surface electromyographic signals are detected through the surface electrodes on the surface of the residual limb muscles, which are collected, amplified and identified. To control the prosthesis. Muscles are composed of muscle cells, which are accompanied by certain bioelectrical effects when contracting autonomously. When the human limbs have to complete certain actions, the corresponding instructions are transmitted from the brain to the muscles in the form of electrical impulses through the spinal cord and motor nerves. At this time, the neuromuscular junction (synapse) undergoes an electrochemical reaction and generates microvolt-level electrical signals. , Causing muscle contraction. The EMG signal control system detects and amplifies the surface EMG signal, recognizes the corresponding action type, and generates control commands for the prosthesis according to the action type. In addition, because the surface EMG signal is the sum of the electrical distribution of many motor units, the waveform is disturbing, and it is difficult to distinguish the waveform of a single unit action from it. The surface EMG signal has the above-mentioned weakness so that the surface EMG signal is easily interfered by external noise sources. Therefore, in order to obtain better results, the surface EMG signal is often mixed with other signals for joint processing. Muscle impedance technology is a kind of bioelectrical impedance technology. Muscle impedance technology uses current electrodes to apply excitation currents (such as high-frequency and low-intensity alternating currents) to the measured muscle tissue area, and analyzes the muscle tissue voltage signals detected by the measurement electrodes. , Extract the impedance characteristics and changes that are closely related to the changes of muscle composition, structural destruction, neuromuscular diseases and other muscle physiological states. During the movement of muscle tissue, the muscle impedance also changes with the muscle tissue. The muscle impedance can be regarded as obtained by the parallel connection of resistance, capacitance and inductance. Among them, since the value of the inductive reactance is very small, it can be ignored. The value under the action is also very small and can be ignored. Therefore, under the action of high-frequency current, the value of muscle impedance is mainly determined by the value of the resistance. When a constant high-frequency current is passed to the muscle tissue, the muscle impedance can be measured indirectly by measuring the voltage difference between the two points in the muscle tissue according to Ohm's law, that is, the muscle impedance signal is obtained.

As mentioned above, in order to make the prosthesis control effect based on EMG signal better, the EMG signal can be combined with other signals for joint processing. Since the muscle impedance signal can also reflect the changes of muscle tissue, it can be combined with the EMG signal Muscle impedance signals are jointly processed to achieve the above objectives. In order to realize the combined processing of EMG signals and muscle impedance signals, it is necessary to use a system that can achieve simultaneous acquisition of EMG signals and muscle impedance signals. The muscle signals described in this embodiment are obtained by using a system that can simultaneously acquire myoelectric signals and muscle impedance signals. Exemplarily, as described above, the system may include an upper computer and a lower computer, and the upper computer and the lower computer are communicatively connected. The lower computer is configured to collect muscle signals. The muscle signals may include myoelectric signals and muscle impedance signals, and send the muscle signals to the upper computer. Machine, the host computer is set to recognize and process muscle signals to generate control commands for the prosthesis. In an embodiment, the lower computer may include an acquisition module and a control module, and the control module is connected to the acquisition module and the upper computer, respectively. The host computer may include a muscle signal acquisition module, an action type recognition module, and a control instruction generation module. The control module is set to control the acquisition module to synchronously acquire the myoelectric signal and the muscle impedance signal, and send the muscle signal and the muscle impedance signal to the host computer. The muscle signal acquisition module may be configured to acquire the muscle signal at the connection of the prosthesis; the action type identification module may be configured to identify the action type corresponding to the muscle signal; the control instruction generation module may be configured to generate the control instruction of the prosthesis according to the action type. In an embodiment, the acquisition module may include an electrode unit and a simulated front-end unit. The simulated front-end unit includes a first channel and a second channel; the electrode unit is disposed on the surface of muscle tissue where the prosthesis is connected, and the electrode unit and the first channel are respectively Connected to the second channel; the analog front-end unit is configured to send the modulated signal to the electrode unit. The electrode unit serves as both an excitation electrode for the high-frequency constant current source and an input electrode for the muscle impedance signal. The first channel is configured to perform the modulation signal The EMG signal is processed, and the second channel is set to process the modulated signal to obtain the muscle impedance signal.

The principle of muscle signal-based prosthetic limb control can be as follows: the information source for controlling the prosthetic limb with muscle signals is the residual muscle group of the residual limb, an electrode unit is placed on the surface of the residual limb muscle, and the control module controls the acquisition module to synchronously collect myoelectric signals and muscle impedance signals , And send the myoelectric signal and muscle impedance signal to the host computer, obtain the muscle signal at the prosthesis connection through the muscle signal acquisition module in the host computer, the action type recognition module recognizes the action type corresponding to the muscle signal and the control instruction generation module according to the action After generating the control commands of the prosthesis, the prosthesis control based on muscle signals is realized.

Compared with the related art in which the prosthetic limbs are controlled only by EMG signals, the beneficial effect of the above setting is that the prosthetic limbs are controlled by the synchronously acquired EMG signals and muscle impedance signals, and the accuracy of prosthetic limb control is improved.

Step 120: Identify the action type corresponding to the muscle signal.

In this embodiment, after acquiring the muscle signal at the joint of the prosthesis, the muscle signal needs to be identified to determine the action type corresponding to the muscle signal, so that the control command of the prosthesis can be generated according to the action type. In an embodiment, the action type may generally include common motion patterns of the prosthesis, such as handshake, roll-up, roll-down, gesture V and gesture OK, etc.

The above process of identifying the action type corresponding to the muscle signal may be pattern recognition. With the continuous development of pattern recognition theory, more and more classifier models are applied to the pattern recognition of muscle signals. The pattern recognition of the muscle signal may be: the acquired muscle signal is input as an input variable into a pre-trained action type recognition model, and the action type corresponding to the muscle signal is obtained through calculation of the action type recognition model. In this embodiment, the The trained classifier model may be generated based on the classifier model training by a set number of training samples, and the training samples may include myoelectric sample signals, muscle impedance sample signals, and action types corresponding to the myoelectric sample signals and muscle impedance sample signals. Commonly used classifier models include Bayesian decision, maximum likelihood classifier, Bayesian classifier, cluster analysis model, neural network model, support vector machine model, chaos and fractal model, and hidden Markov model Model etc. In this embodiment, the classifier model can be set according to the actual situation, which is not limited herein. The following uses cluster analysis model and neural network model as examples to illustrate.

The basic idea of the cluster analysis model is to classify according to the similarity of multiple pattern features to be classified, similar to one category, and dissimilar to another. To put it simply, similarity means that multiple components between two feature vectors are relatively close. Cluster analysis includes two basic contents, namely the measurement of pattern similarity and clustering algorithm. The cluster analysis model is a linear classification model. The algorithm of the cluster analysis model is simple and the operation speed is fast. In the pattern classification used for muscle signals, the overlapping areas of different categories are small and the difference is obvious. Using this model can avoid Many non-linear classifications have a lengthy training process, resulting in better recognition results.

The full name of the neural network is Artificial Neural Networks (ANN). ANN is based on the basic principles of neural networks in biology. After understanding and abstracting the human brain structure and external stimulus response mechanism, the network topology knowledge is used as the theoretical basis , A mathematical model that simulates the processing of complex information by the nervous system of the human brain. The model relies on the complexity of the system and adjusts the weights of the interconnection between a large number of internal nodes (neurons) to process information. The ANN model has the advantages of self-learning, self-adaptation, self-organization, nonlinearity and deep parallel operation. The ANN model consists of a large number of nodes (or neurons) connected to each other. Each node represents a specific output function, called an activation function. The connection between each two nodes represents a weight for the signal passing through the connection. The output of the ANN model depends on the structure of the network, the connection method of the network, the weight and the activation function. Taking the ANN model as a three-layer structure as an example, the ANN model includes an input layer, a hidden layer, and an output layer. The input weight of the ANN model is the weight of the input node of the ANN model to the hidden layer node; the threshold of the ANN model is the threshold of the hidden layer node. It should also be noted that the input weights of the ANN model and the threshold of the ANN model are randomly set. There are two ways to train a set number of training samples based on the ANN model to generate a pre-trained sports type recognition model:

Method 1: Obtain EMG sample signal, muscle impedance sample signal and the action type corresponding to EMG sample signal and muscle impedance sample signal; superimpose EMG sample signal and muscle impedance sample signal to obtain superimposed muscle sample signal; superimpose muscle sample The signal is used as the input variable of the ANN model, and the action types corresponding to the EMG sample signal and the muscle impedance sample signal are used as the output variable of the ANN model, and the ANN is determined according to the input variable, the output variable, the input weight of the ANN model and the threshold value of the ANN model The output weight of the model; determine the pre-trained sports type model according to the input weight, threshold and output weight of the ANN model.

Method 2: Obtain the EMG sample signal, muscle impedance sample signal and the action type corresponding to the EMG sample signal and the muscle impedance sample signal; extract the characteristics of the EMG sample signal to obtain the EMG sample characteristic signal, and extract the muscle impedance sample signal The characteristic of the muscle impedance sample is obtained; the characteristic signal of the electromyography sample and the characteristic signal of the muscle impedance sample are superimposed to obtain the characteristic signal of the superimposed muscle sample; the characteristic signal of the superimposed muscle sample is used as the input variable of the ANN model, and the signal of the electromyography sample and the muscle are used The action type corresponding to the impedance sample signal is used as the output variable of the ANN model. The output weight of the ANN model is determined according to the input variable, the output variable, the input weight of the ANN model and the threshold of the ANN model; according to the input weight, threshold and The output weights determine the pre-trained sports type model.

The types of movements can generally include common movement patterns of prosthetics, such as handshake, roll-up, roll-down, gesture V and gesture OK.

In an embodiment, on the basis of the above technical solution, identifying the action type corresponding to the muscle signal may include: superimposing the myoelectric signal and the muscle impedance signal to obtain the superimposed muscle signal. The superimposed muscle signal is input into a pre-trained muscle action type recognition model to obtain the action type corresponding to the muscle signal.

In an embodiment, identifying the action type corresponding to the muscle signal may include: superimposing the myoelectric signal and the muscle impedance signal to obtain a superimposed muscle signal, and inputting the superimposed muscle signal as an input variable to a pre-trained muscle action type recognition model In the calculation of the muscle action type recognition model, the action type corresponding to the muscle signal is obtained.

The above-mentioned superposition of myoelectric signals and muscle impedance signals to obtain superimposed muscle signals, and input the superimposed muscle signals as input variables to a pre-trained muscle action type recognition model to obtain the action type corresponding to the muscle signal, thereby achieving signal-level pattern recognition .

In an embodiment, on the basis of the above technical solution, a muscle action type recognition model can be trained by acquiring the EMG sample signal, the muscle impedance sample signal, and the action type corresponding to the EMG sample signal and the muscle impedance sample signal ; Superimpose the EMG sample signal and the muscle impedance sample signal to obtain the superimposed muscle sample signal; use the superimposed muscle sample signal as the input variable, the action type corresponding to the EMG sample signal and the muscle impedance sample signal as the output variable, and train the classifier model to obtain Muscle movement type recognition model.

In one embodiment, the muscle action recognition model can be trained by acquiring the EMG sample signal, the muscle impedance sample signal and the action type corresponding to the EMG sample signal and the muscle impedance sample signal, the above constitutes a set number of groups of training sample. In order to improve the recognition rate of the action type of the muscle action recognition model obtained by training, the following two requirements are required for the selection of training samples: First, it can be understood that multiple actions of the human body are completed by multiple muscle tissues. Coordinated and completed together. During the completion of the action, the participation time of multiple muscle tissues is different, the contribution to the completion of the action is also different, and the resulting muscle signals are also different. The muscle signals of different muscle tissues of the same subject are different, the muscle signals of the same muscle tissue of different subjects are also different, and the muscle signals of the same muscle tissue of the same subject under different action types are also different. . In order to make the recognition rate of the muscle action recognition model obtained by training higher, it is necessary to minimize the influence of the choice of measurement site on the pattern recognition rate. On the basis of studying the correspondence between common action types and multiple muscle groups, the muscle groups that play a leading role under each action type can be selected to use the muscle signals of these muscle groups to complete the identification of corresponding movement types; Second, after the determination of the muscle signal measurement site is completed, the action type needs to be selected to minimize the impact of the action type selection on the pattern recognition rate, mainly based on the following two principles: First, the desired There is a direct correspondence between the action type and the measurement part to ensure that the muscle signal amplitude of the measurement part under this action type is large and the signal is obvious; Second, the action type to be selected needs to be a common exercise type in daily life to ensure the cost The practicality of the embodiment, after the establishment of the muscle movement type recognition model, enables the muscle movement type recognition model to be applied to the actual development of prostheses. In this embodiment, on the basis of meeting the above two requirements for the selection of training samples, the selection can be made according to the actual situation, which is not limited herein. Exemplary, for example, the radial wrist flexors and ulnar wrist extensors are selected as the muscle tissue for extracting muscle signals. This is because the above two muscle tissues are closely related to the common movement types of the arm. The movement of the arm is in these two muscles The generated muscle signals are more obvious, and the two muscle tissues are larger in size, so the muscle signals collected will not be affected by the muscle signals of other adjacent muscle tissues. On this basis, select the type of movement: fist, flip, flip, gesture V and gesture OK. Based on the above, the muscle signals of the measurement sites under these five exercise types are extracted to constitute a set number of training samples.

In an embodiment, the set number of training samples may be a number group formed from training samples of the same subject, or may be a number group formed from training samples of different subjects. The situation is set, and it is not limited here. In an embodiment, considering the difference of the aforementioned muscle signals, the set number of training samples is formed from the training samples of the same subject. The beneficial effect of this setting is that it can be established based on this The prediction accuracy of the muscle movement type recognition model is better. In this embodiment, the classifier model can also be set according to the actual situation, which is not limited here.

After obtaining a set number of training samples, superimpose the EMG sample signal and the muscle impedance sample signal to obtain the superimposed muscle sample signal, and use the superimposed muscle sample signal as the input variable. The actions corresponding to the EMG sample signal and the muscle impedance sample signal The type is used as an output variable, and the classifier model is trained to obtain a muscle action type recognition model. The following uses the classifier model as the ANN model as an example to further explain the training of the classifier model to obtain the muscle action type recognition model, specifically:

Exemplarily, if M sets of training samples (x _i , y _i ) are set, where x _i represents the i-th input variable, x _i is an n-dimensional vector; y _i is the i-th output variable, and y _i is m-dimensional vector; ω _i is the input weight of the i-th input node and the hidden layer node; θ _i is the threshold of the i-th hidden layer node; β _i is the i-th output weight. There are N hidden layer nodes, g(x) is the activation function, then the ANN model is

Among them, j = 1, 2, ..., M. The input variable x _i , the input weight ω _{i of the} ANN model and the threshold θ _i of the ANN model

Perform training to determine the output weight β _i . Then, based on the input weight ω _i of the ANN model, the threshold θ _{i of the} ANN model, and the output weight β _{i of the} ANN model, a pre-trained muscle action type recognition model is determined. When n=4 and m=5, the ANN model designed in this embodiment is that the input layer has 4 neuron nodes, corresponding to the dimension of the input vector t. The selection of the number of neuron nodes in the output layer should be determined according to the dimension of the output variable y _i , that is, the number of motion types. The output layer of t neuron nodes can combine 2 ^t output sequences and 2 ^t output sequences. The output motion type corresponds to. Since the number of motion types is 5, the number of neurons in the output layer is set to 3, which can complete the recognition of 5 motion types. There is no exact theoretical basis for determining the number of neurons in the hidden layer, but there are references for experience, and the number of neuron nodes in the hidden layer can be determined after repeated experiments. For example, the number of neuron nodes in the hidden layer is set to 6. So far, the pre-set ANN model is trained to obtain a muscle action type recognition model with 4 neural nodes in the input layer, 6 neural nodes in the hidden layer and 3 neuron nodes in the output layer.

In an embodiment, on the basis of the above technical solution, identifying the action type corresponding to the muscle signal may include: extracting the characteristics of the EMG signal to obtain the EMG characteristic signal, and extracting the characteristics of the muscle impedance signal to obtain the muscle impedance characteristic signal ; Superimposing the myoelectric characteristic signal and the muscle impedance characteristic signal to obtain the superimposed characteristic signal; input the superimposed characteristic signal into the pre-trained characteristic action type recognition model to obtain the action type corresponding to the muscle signal.

In an embodiment, identifying the action type corresponding to the muscle signal may include: extracting the characteristics of the EMG signal to obtain the EMG characteristic signal, and extracting the characteristics of the muscle impedance signal to obtain the muscle impedance characteristic signal; superimposing the EMG characteristic signal and the muscle The impedance characteristic signal is used to obtain the superimposed characteristic signal; the superimposed characteristic signal is input into the pre-trained characteristic action type recognition model to obtain the action type corresponding to the muscle signal, and the action type corresponding to the muscle signal is obtained through calculation of the characteristic action type recognition model.

Extracting the characteristics of the EMG signal to obtain the EMG characteristic signal, and extracting the characteristics of the muscle impedance signal to obtain the muscle impedance characteristic signal, superimposing the EMG characteristic signal and the muscle impedance characteristic signal to obtain the superimposed characteristic signal, and using the superimposed characteristic signal as The input variable is input to the pre-trained feature action type recognition model to obtain the action type corresponding to the muscle signal, and the feature level pattern recognition is realized.

In an embodiment, on the basis of the above technical solution, a characteristic action type recognition model may be trained by acquiring the EMG sample signal, the muscle impedance sample signal, and the action type corresponding to the EMG sample signal and the muscle impedance sample signal ; Extract the characteristics of EMG sample signals to obtain EMG sample characteristic signals, and extract the characteristics of muscle impedance sample signals to obtain muscle impedance sample characteristic signals; superimpose EMG sample characteristic signals and muscle impedance sample characteristic signals to obtain superimposed muscle sample characteristic signals The superimposed muscle sample feature signal is used as an input variable, the action type corresponding to the EMG sample signal and the muscle impedance sample signal is used as the output variable, and the classifier model is trained to obtain a characteristic action type recognition model.

In one embodiment, feature extraction (also called dimensionality reduction or dimensionality reduction) refers to mapping or transforming the original training data from the original high-dimensional data space to the low-dimensional feature space to obtain the most reflective of the essence of the training data Low-dimensional data features. Exemplarily, as feature extraction can be described as a process, a transformation is performed on the feature vector U = (u ₁ , u ₂ , ..., u _Q ) ^T : v _s = h _s (U), s = 1, 2,...,P,P<Q, resulting in a dimension-reduced feature vector V=(v ₁ , v ₂ ,..., v _P ) ^T. The quality of the extracted low-dimensional data features has a greater impact on the accuracy of pattern recognition. In order to improve the accuracy of pattern recognition, it is necessary to extract features from the acquired EMG sample signal and muscle impedance sample signal. The feature extraction of EMG sample signal is to extract the most effective and most distinguishable component of EMG sample signal from other EMG types from the collected EMG sample signal. Similarly, the feature extraction of the muscle impedance sample signal is to extract the most effective and most distinguishable component of the muscle impedance sample signal from the collected muscle impedance sample signal under other action types. For EMG sample signals, feature extraction can use time domain analysis, frequency domain analysis, time-frequency domain analysis, and chaotic fractal method. Among them, when the time domain analysis method extracts the characteristics of the EMG sample signal, the EMG sample signal is usually regarded as a random signal with one-dimensional zero mean and variance that varies with the intensity of the action. When performing time domain analysis, the most obvious feature of the EMG sample signal is the amplitude of the EMG sample signal, followed by other feature quantities such as the variance and mean square error of the EMG sample signal. The time-domain features of EMG sample signals have the advantages of easy extraction and simple calculation, but small changes in muscle contraction force can cause large fluctuations in time-domain feature elements such as mean or variance, so use EMG sample signals The time-domain features of pattern recognition for action types have certain randomness and instability. The frequency domain analysis method can usually use power spectrum estimation. The main frequency domain characteristics of power spectrum estimation include parameters such as frequency range, power spectrum maximum value, power spectrum maximum frequency, median frequency and average power frequency. Because the Fourier transform can only convert the time domain signal to the frequency domain, during the conversion process, any information in the time domain of the non-stationary signal is lost, and only the frequency domain characteristics of the signal are retained. However, the EMG sample signal is a time-varying non-stationary signal. Feature extraction of the EMG sample signal based on the time domain or frequency domain method will lose part of the characteristics of the EMG sample signal. Based on the above, the time-frequency domain analysis method for non-stationary signals can be used. Common time-frequency domain analysis methods include wavelet transform, wavelet packet transform, short-time Fourier transform, and Wigner distribution. Of course, it can be understood that the feature extraction method can be selected according to the actual situation, which is not limited herein.

The classifier model can also choose the ANN model. The training process of the characteristic action type recognition model can refer to the above-mentioned muscle movement type recognition model. The difference lies in the different input variables, but the two ideas are the same, and will not be repeated here.

In an embodiment, the classifier model used in training the feature action type recognition model may be the same as the classifier model used in training the muscle movement type recognition model, if both are ANN models, or different, such as training feature action type The classifier model used when identifying the model is the ANN model, and the classifier model used when training the muscle movement type identification model is the cluster analysis model. It can be set according to the actual situation and is not limited here.

In an embodiment, extracting the features of the EMG signal to obtain the EMG feature signal, and extracting the features of the muscle impedance signal to obtain the muscle impedance feature signal can use the same feature extraction method as when training the feature action type recognition model.

In an embodiment, on the basis of the above technical solution, identifying the action type corresponding to the muscle signal may include: inputting the EMG signal into a pre-trained EMG action type recognition model to obtain the EMG action type; The impedance signal is input into a pre-trained muscle impedance action type recognition model to obtain the muscle impedance action type; the action type corresponding to the muscle signal is determined according to the myoelectric action type and the muscle impedance action type.

In an embodiment, identifying the action model corresponding to the muscle signal may include: inputting the EMG signal into the pre-trained EMG action type recognition model to obtain the EMG action type; and inputting the muscle impedance signal to the pre-trained muscle In the impedance action type recognition model, the muscle impedance action type is obtained. Through the above process, based on the myoelectric signal and the muscle impedance signal to obtain the myoelectric action type and muscle impedance action type recognition results, the action type corresponding to the muscle signal can be determined according to the myoelectric action type and the muscle impedance action type. Here, the determination of the action type corresponding to the muscle signal according to the type of myoelectric action and the type of muscle impedance action can be understood as follows: the proportion of the type of myoelectric action and the type of muscle impedance action is preset, and the action type corresponding to the muscle signal is determined according to the ratio. In an embodiment, the above ratio can be determined in the training myoelectric action type recognition model and the muscle impedance action type recognition model.

The above input EMG signal to the pre-trained EMG action type recognition model to obtain the EMG action type, and the muscle impedance signal to the pre-trained muscle impedance action type recognition model to obtain the muscle impedance action type, according to the EMG The action type and muscle impedance action type determine the action type corresponding to the muscle signal, and realize the pattern recognition at the decision level.

In one embodiment, on the basis of the above technical solution, the EMG action type recognition model can be trained in the following manner: acquiring the EMG sample signal and the action type corresponding to the EMG sample signal; using the EMG sample signal as the input variable , The action type corresponding to the EMG sample signal is used as the output variable, and the classifier model is trained to obtain the EMG action type recognition model.

The muscle impedance action type recognition model can be trained by acquiring the muscle impedance sample signal and the action type corresponding to the muscle impedance sample signal; using the muscle impedance sample signal as the input variable, and the action type corresponding to the muscle impedance sample signal as the output variable, training The classifier model is used to obtain the muscle impedance action type recognition model.

In an embodiment, the classifier model used in training the EMG action type recognition model may be an ANN model, and the classifier model used in training the muscle impedance action type recognition model may also be an ANN model, EMG action type recognition model. The training process of the muscle impedance action type recognition model can refer to the aforementioned muscle movement type recognition model, the difference is that the input variables are different, but the two ideas are the same, and will not be repeated here.

In an embodiment, the classifier model used when training the electromyography action type recognition model, the classifier model used when training the muscle impedance action type recognition model, the classifier model used when training the characteristic action type recognition model, and the training muscle The classifier model used in the motion type recognition model can be the same, such as the ANN model, or it can be different, such as the classifier model used in training the EMG action type recognition model and the classifier model used in the training feature action type recognition model Both are ANN models, and the classifier model used when training the muscle impedance action type recognition model and the classifier model used when training the muscle movement type recognition model are cluster analysis models. It can be set according to the actual situation and is not limited here.

Step 130: Generate a control instruction for the prosthesis according to the action type.

In this embodiment, after identifying the corresponding action type according to the muscle signal, a control instruction for the prosthesis can be generated according to the action type, and the control instruction can be used to instruct the driving module to drive the prosthesis to perform the corresponding action. In one embodiment, it is also possible to control the progress of related operations involving the prosthesis.

Exemplarily, if the corresponding motion type is recognized as a handshake based on the muscle signal, the control command that can generate the prosthesis according to the motion type is a handshake. As another example, if a prosthesis participates in a Bluetooth car experiment, the Bluetooth car includes a Bluetooth communication module, a controller, a remote control, and a motor drive module. The remote control interface is provided with buttons for forward, backward, left turn, right turn, and stop. The working principle of the Bluetooth car is that the Bluetooth communication module is paired with the remote control's Bluetooth to receive the action command sent by the remote control. The action command can be generated by pressing the corresponding button on the remote control interface and send the action command to the controller , The controller analyzes the action command to control the motor drive module to drive the motor forward and reverse to achieve the car's forward, backward, left, right, or stop. Now the remote control can be controlled by the prosthetic limb, the corresponding action type can be identified according to the muscle signal, the control command of the prosthetic limb can be generated according to the action type, the prosthetic limb can be controlled according to the control command to control the remote control to generate the corresponding motion instruction, and the motion instruction can be sent through the Bluetooth communication module To the controller, the controller analyzes the received motion command and controls the motor drive module to drive the motor forward or reverse, so as to realize the forward, backward, left, right or stop of the car.

In the technical solution of this embodiment, by acquiring muscle signals at the joint of the prosthesis, the muscle signals include the electromyography signal and the muscle impedance signal, the action type corresponding to the muscle signal is identified, and the control commands for the prosthesis are generated according to the action type, thereby improving the accuracy of prosthesis control rate.

In an embodiment, based on the above technical solution, the EMG signal is obtained after processing according to the muscle impedance signal.

In an embodiment, during the process of collecting EMG signals, there may be a situation where the electrodes fall off. At this time, the obtained EMG signals are essentially interference signals, that is, motion artifacts of the EMG signals. And because the electrode falls off, the acquired muscle impedance signal will be very large, such as greater than the preset threshold. Based on the above, in this case, it can be determined whether the acquired EMG signal is an interference signal according to the muscle impedance signal, and if the acquired EMG signal is determined to be the interference signal according to the muscle impedance signal, the EMG signal can be used as Discard processing, that is, the EMG signal is an invalid EMG signal. If it is determined that the acquired EMG signal is not an interference signal according to the muscle impedance signal, the EMG signal will be retained, that is, the EMG signal is a valid EMG signal, and can be used for subsequent data processing.

The above determines whether the EMG signal is an interference signal based on the muscle impedance signal. If the EMG signal is determined to be the interference signal, the EMG signal is discarded to achieve the elimination of motion artifacts of the EMG signal, thereby improving pattern recognition Accuracy.

In one embodiment, based on the above technical solution, the EMG signal is obtained after processing the muscle impedance signal, which may include: acquiring the EMG signal to be screened, the EMG signal to be screened and the muscle impedance signal are synchronized Obtained; if the muscle impedance signal is less than the preset threshold, the EMG signal to be screened is used as the EMG signal.

In an embodiment, the preset threshold can be used as a criterion for determining whether the EMG signal to be filtered is an interference signal. It can be understood that the EMG signal included in the muscle signal is obtained after the muscle impedance signal is processed. Before that, the EMG signal to be screened is obtained, and the EMG signal and the muscle impedance signal to be screened are acquired synchronously If the muscle impedance signal is less than the preset threshold, it can indicate that the EMG signal to be screened is not an interference signal. In other words, the EMG signal to be screened is a valid EMG signal. As an EMG signal. If the muscle impedance signal is greater than or equal to the preset threshold, it can indicate that the EMG signal to be screened is an interference signal. In other words, the EMG signal to be screened is an invalid EMG signal, at this time, the EMG to be screened The signal is discarded.

In this embodiment, since the EMG signal and the muscle impedance signal to be screened are collected synchronously, and the EMG signal belongs to the EMG signal to be screened, it can be understood that the EMG signal and the muscle impedance signal are also collected synchronously owned.

The EMG signals described in this embodiment are all obtained after processing according to the muscle impedance signal. Similarly, in the training stage, the EMG sample signals are all obtained after processing according to the muscle impedance sample signal.

Example 2

FIG. 2 is a schematic structural diagram of a prosthetic limb control device provided by an embodiment. This embodiment can be applied to a case where the prosthesis is performed based on myoelectric signals and muscle impedance signals. The device can be implemented in software and/or hardware. The device can be configured in a device, such as a computer. As shown in FIG. 2, the device includes: a muscle signal acquisition module 210 configured to acquire a muscle signal at a prosthetic junction, the muscle signal includes an electromyographic signal and a muscle impedance signal; an action type recognition module 220 is configured to identify a muscle signal corresponding to Action type; the control instruction generation module 230 is configured to generate a control instruction of the prosthesis according to the action type.

In an embodiment, based on the above technical solution, the action type recognition module 220 may include: a superimposed muscle signal acquisition unit configured to superimpose the electromyography signal and the muscle impedance signal to obtain the superimposed muscle signal; the first recognition of the action type The unit is configured to input the superimposed muscle signal into the pre-trained muscle action type recognition model to obtain the action type corresponding to the muscle signal.

In an embodiment, on the basis of the above technical solution, the action type recognition module 220 may include: a characteristic signal acquisition unit configured to extract the characteristics of the EMG signal to obtain the EMG characteristic signal, and extract the characteristics of the muscle impedance signal Obtain the muscle impedance characteristic signal; the superimposed characteristic signal acquisition unit is set to superimpose the myoelectric characteristic signal and the muscle impedance characteristic signal to obtain the superimposed characteristic signal; the action type second recognition unit is set to input the superimposed characteristic signal to the pre-trained characteristic action In the type recognition model, the action type corresponding to the muscle signal is obtained.

In an embodiment, based on the above technical solution, the action type recognition module 220 may include: an electromyography action type acquisition unit configured to input an electromyography signal into a pre-trained electromyography action type recognition model to obtain Muscle action type; Muscle impedance action type acquisition unit, set to input muscle impedance signal into a pre-trained muscle impedance action type recognition model to obtain muscle impedance action type; Action type third recognition unit, set to be based on myoelectric action The type and muscle impedance action type determine the action type corresponding to the muscle signal.

In an embodiment, based on the above technical solution, the above-mentioned device further includes a muscle action type recognition model training module, which is configured to train a muscle action type recognition model in the following manner: acquiring EMG sample signals, muscle impedance sample signals and The action type corresponding to the EMG sample signal and the muscle impedance sample signal; superimposing the EMG sample signal and the muscle impedance sample signal to obtain the superimposed muscle sample signal; using the superimposed muscle sample signal as the input variable, the EMG sample signal corresponds to the muscle impedance sample signal The action type is used as an output variable, and the classifier model is trained to obtain a muscle action type recognition model.

In an embodiment, based on the above technical solution, the above-mentioned device further includes a feature action type recognition model training module, which is configured to train the feature action type recognition model by: acquiring the electromyographic sample signal, the muscle impedance sample signal and the The action type corresponding to the EMG sample signal and the muscle impedance sample signal; extracting the characteristics of the EMG sample signal to obtain the EMG sample characteristic signal, and extracting the characteristics of the muscle impedance sample signal to obtain the muscle impedance sample characteristic signal; superimposing the EMG sample characteristic signal And the muscle impedance sample characteristic signal to obtain the superimposed muscle sample characteristic signal; the superimposed muscle sample characteristic signal is used as the input variable, and the action types corresponding to the EMG sample signal and the muscle impedance sample signal are used as the output variable, and the classifier model is trained to obtain the characteristic action type Identify the model.

In an embodiment, on the basis of the above technical solution, the above-mentioned device further includes an EMG type recognition model training module, which is configured to train the EMG type recognition model by: acquiring EMG sample signals and EMG sample signals The action type corresponding to the signal; using the EMG sample signal as the input variable and the action type corresponding to the EMG sample signal as the output variable, the classifier model is trained to obtain the EMG action type recognition model.

In an embodiment, the above device further includes a muscle impedance action type recognition model training module, which is configured to train a muscle impedance action type recognition model by acquiring a muscle impedance sample signal and an action type corresponding to the muscle impedance sample signal; The impedance sample signal is used as an input variable, and the action type corresponding to the muscle impedance sample signal is used as an output variable. The classifier model is trained to obtain a muscle impedance action type recognition model.

In an embodiment, based on the above technical solution, the EMG signal is obtained after processing the muscle impedance signal, which may include: acquiring the EMG signal to be screened, the EMG signal to be screened and the muscle impedance signal It is obtained through synchronous acquisition; if the muscle impedance signal is less than the preset threshold, the EMG signal to be screened is used as the EMG signal.

The prosthetic limb control device provided by this embodiment can execute the prosthetic limb control method provided by any embodiment of the present disclosure, and has corresponding function modules and beneficial effects of the execution method.

Example Three

FIG. 3 is a schematic structural diagram of a prosthetic limb control system provided by an embodiment. This embodiment can be applied to a case where a prosthetic limb is performed based on myoelectric signals and muscle impedance signals. As shown in FIG. 3, the prosthetic limb control system may include: The upper computer 1 and the lower computer 2, the upper computer 1 is provided with the prosthetic limb control device described in the embodiment of the present disclosure, and the upper computer 1 and the lower computer 2 are communicatively connected. The structure and function of the prosthetic limb control system are described below.

The lower computer 2 is configured to collect muscle signals. The muscle signals include myoelectric signals and muscle impedance signals, and send the muscle signals to the upper computer 1.

In this embodiment, the lower computer 2 is configured to collect muscle signals. The muscle signals include myoelectric signals and muscle impedance signals, and send the muscle signals to the upper computer 1. The upper computer 1 is provided with the prosthetic control described in the embodiments of the present disclosure. The device is configured to recognize the muscle signal to obtain a corresponding action type, and generate a control command for the prosthesis according to the action type. The lower computer 2 described here can realize the collection of myoelectric signals and muscle impedance signals.

As shown in FIG. 4, a schematic structural diagram of the main window operation interface of the host computer 1 is given. The main window operation interface (hereinafter referred to as the main interface) can be designed using Matlab application (Application, APP) design (Designer). The main interface may include a signal acquisition module, an acquisition end configuration module, and a mechanical control module, where the signal acquisition module may be configured to display the acquired EMG signals and muscle impedance signals in real time on the waveform display panel and provide coordinate axis display Range adjustment function, real-time filtering and baseline drift correction function, subject basic information creation and modification function, and data saving function. The collection terminal configuration module can be set to provide the working parameter setting function of the collection terminal (ie, the lower computer 2), such as setting the sampling frequency of the lower computer 2, and can be set to display the working status of the collection terminal. The mechanical control module can be set to provide control parameter settings and mechanical real-time control functions. The multiple modules involved in the above main interface and the functions provided by the multiple modules can be realized by clicking the corresponding buttons on the main interface.

In the technical solution of this embodiment, the upper computer receives the muscle signal at the connection of the prosthesis collected by the lower computer. The muscle signal includes an electromyographic signal and a muscle impedance signal, recognizes the action type corresponding to the muscle signal, and generates a control command for the prosthesis according to the action type. Improve the accuracy of prosthetic control.

In an embodiment, as shown in FIG. 3, based on the above technical solution, the lower computer 2 may include an acquisition module 21 and a control module 22, and the control module 22 is connected to the acquisition module 21 and the upper computer 1 respectively. The control module 22 is configured to control the collection module 21 to simultaneously collect the myoelectric signal and the muscle impedance signal, and send the myoelectric signal and the muscle impedance signal to the host computer 1.

In an embodiment, the lower computer 2 may include an acquisition module 21 and a control module 22, and the control module 22 may be configured to control the acquisition module 21 to simultaneously acquire myoelectric signals and muscle impedance signals. In an embodiment, the control module 22 may select the STM8L151F3 chip.

In an embodiment, as shown in FIG. 3, based on the above technical solution, the acquisition module 21 may include an electrode unit 211 and an analog front-end unit 212, and the analog front-end unit 212 may include a first channel and a second channel; an electrode unit 211 is placed on the surface of the muscle tissue where the prosthesis is connected, and the electrode unit 211 is connected to the first channel and the second channel respectively. The analog front-end unit 212 is configured to send the modulated signal to the electrode unit 211, the first channel is configured to process the modulated signal to obtain a myoelectric signal, and the second channel is configured to process the modulated signal to obtain a muscle impedance signal.

In an embodiment, the acquisition module 21 may include an electrode unit 211 and an analog front-end unit 212. The electrode unit 211 may include at least two electrodes. The pole analog front-end unit 212 may include a first channel and a second channel. May include a modulator, the first channel may include a low-pass filter, a first programmable gain amplifier, a first analog-to-digital converter, and a register, and the second channel may include a high-pass filter, a second programmable gain amplifier, a demodulator , The second analog-to-digital converter and registers.

The electrode is a kind of sensor. The role of the electrode is to convert the displacement current of ion conduction in the human body into the conduction current of electronic conduction in the detection circuit. As can be seen from the foregoing, the EMG signal described in the embodiments of the present disclosure refers to the surface EMG signal, which is measured by using the surface electrode as a guide electrode, and the muscle impedance signal can be obtained indirectly by measuring the voltage difference between two points Since electrodes are attached to the muscle surface of the prosthesis when measuring EMG signals, the EMG muscle impedance signal can be collected through the common electrode. There are four-electrode method and two-electrode method for measuring muscle impedance signals using electrodes. Among the four-electrode method, two of the four electrodes are used as the output electrode of the excitation signal, and the other two are used as the input electrode of the muscle impedance signal; The two electrodes are used as both the output electrode of the excitation signal and the input electrode of the muscle impedance signal. The above can be selected according to the actual situation and is not limited here. In an embodiment, a two-electrode method is used, that is, a pair of electrodes is used to simultaneously acquire myoelectric signals and muscle impedance signals. At this time, the electrode unit 211 includes two electrodes. In order to achieve simultaneous acquisition of EMG and muscle impedance signals, the EMG and muscle impedance signals need to be placed in different frequency bands. The modulator may be configured to output a modulated signal and send the modulated signal to the electrode unit 211. The modulated signal may be used as an excitation signal for measuring the muscle impedance signal. The excitation signal is applied to the surface of the muscle tissue placed at the joint of the prosthesis. On the two electrodes, the frequency of the excitation signal is usually much higher than the frequency band of the EMG signal. Based on the above, the electrode unit 211 collects a mixed signal including the EMG signal and the muscle tissue signal. The mixed signal is filtered by a low-pass filter of the first channel to remove a high-frequency signal to obtain a low-frequency signal including an EMG signal, and then enters a register for storage through a first programmable gain amplifier and a first analog-to-digital converter. At the same time, the mixed signal passes through the high-pass filter of the second channel to filter out the low-frequency signal to obtain a high-frequency modulated signal containing the muscle impedance signal, and then passes through the second programmable gain amplifier to input the modulated signal to the demodulator for demodulation to obtain the corresponding Voltage signal, the corresponding muscle impedance signal is obtained according to the voltage signal, and the muscle impedance signal enters the register for storage via the second analog-digital converter.

In an embodiment, the analog front-end unit 212 may use an ADS1292R chip.

In an embodiment, based on the above technical solution, the lower computer 2 may further include a Bluetooth module, the Bluetooth module is connected to the control module 22 and the upper computer 1 respectively, and the Bluetooth module is configured to implement communication between the control module 22 and the upper computer 1 connection.

In an embodiment, the upper computer 1 is also provided with a Bluetooth module. The Bluetooth module in the upper computer 1 can be set to the master mode, and the Bluetooth module in the lower computer 2 can be set to the slave mode. Both are configured with the same baud To achieve a communication connection between the control module 22 and the host computer 1. In an embodiment, in order to ensure the consistency of data transmission, the baud rate needs to be consistent with the baud rate set by the control module and the acquisition module. In one embodiment, the Bluetooth module can choose HJ-580X wireless Bluetooth serial port transparent transmission module, which has the advantages of small size and low power consumption, and supports serial port transparent transmission and can change the master and slave of the module through commands, communication The distance can be up to 20 meters (m).

In an embodiment, on the basis of the above technical solution, the lower computer 2 may further include a power module, the power module is respectively connected to the acquisition module 21, the control module 22 and the Bluetooth module and is the acquisition module 21, the control module 22 and the Bluetooth module Provide electrical energy.

In one embodiment, in order to reduce power frequency interference and enhance the portability of the lower computer 1, the power module may use a 3.7 volt (V) lithium battery, while using a complementary metal oxide semiconductor (Complementary Metal Oxide Semiconductor (CMOS) low-voltage drop stability Voltage protector for circuit protection.

The following uses the analog front-end unit 211 to select the ADS1292R chip and the control module 22 to select the STM8L151F3 chip as an example to explain the working principle between the control module 22 and the acquisition module 21.

The ADS1292R chip is a chip for bioelectric signal acquisition and analog-to-digital conversion launched by TI. It is suitable for the measurement of myoelectric signals and muscle impedance signals. The ADS1292R chip has many features required in applications such as portability, low-power medical and motion monitoring. The ADS1292R chip includes two low-noise programmable gain amplifiers (Pmgrammable Gain Amplifier, PGA) and two 24-bit high-resolution analog-to-digital converters (Analog to Digital Converter, ADC), which can achieve EMG signals and muscles Simultaneous acquisition of impedance signals and simultaneous conversion. As shown in FIG. 5, a schematic structural diagram of the analog front-end unit 211 is given. To meet actual needs, the electrode unit 211 is not directly connected to the analog front-end module 212, but a low-pass filter and a high-pass filter are provided between the electrode unit 211 and the analog front-end module 212. Among them, channel 1 collects myoelectric signals, and channel 2 collects muscle impedance signals. The STM8L151F3 chip is an ultra-low power chip designed for high coding efficiency and performance. The STM8L151F3 chip is set to control the operation of the analog front-end unit 212, the Bluetooth module and the power module. The STM8L151F3 chip is the functional link between the analog front-end unit 212 and the host computer 1. On the one hand, the STM8L151F3 chip passes the serial peripheral interface (Serial Peripheral Interface, SPI) bus Control the registers of the ADS1292R chip to realize the simultaneous acquisition of EMG signals and muscle impedance signals. On the other hand, the STM8L151F3 chip communicates with the Bluetooth module through the serial port, receives the command of the host computer 1, and controls the Bluetooth module to collect the collected EMG signals and The muscle impedance signal is uploaded to the host computer 1. The following describes the acquisition process of myoelectric signal and muscle impedance signal.

The ADS1292R chip modulator outputs a 32 kilohertz (kHz) AC square wave signal as the excitation signal for measuring muscle impedance signals. The frequency of the square wave signal is much higher than the frequency band of the EMG signal. At this time, the differential electrode (ie electrode 1 and The electrode 2) acquires a mixed signal containing myoelectric signal and muscle impedance signal. This mixed signal uses a low-pass filter in channel 1 to filter out high-frequency signals to obtain low-frequency components containing myoelectric signals; channel 2 uses a high-pass filter to filter out low-frequency signals to obtain high-frequency modulated signals including muscle impedance signals, and then passes The demodulator inside the ADS1292R chip demodulates to obtain the corresponding voltage signal, and the corresponding muscle impedance signal can be calculated from the voltage signal.

In this embodiment, the analog front-end unit 211, the control module 22, the Bluetooth module, and the power supply module can all be type-selected according to actual conditions, which is not limited herein.

Example 4

6 is a schematic structural diagram of a device provided by an embodiment. 6 shows a block diagram of an exemplary device 412 suitable for implementing embodiments of the present disclosure. The device 412 shown in FIG. 6 is only an example, and should not bring any limitation to the functions and use scope of the embodiments of the present disclosure.

As shown in FIG. 6, the device 412 is represented in the form of a general-purpose computing device. The components of device 412 may include, but are not limited to, one or more processors 416, system memory 428, and bus 418 connected to different system components (including system memory 428 and processor 416).

The bus 418 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus that uses any of a variety of bus structures. For example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, enhanced ISA bus, Video Electronics Standards Association (VideoElectronicsStandardsAssociation) , VESA) local bus and peripheral component interconnect (Peripheral Component Interconnect, PCI) bus.

Device 412 includes a variety of computer system readable media. These media may be any available media that can be accessed by device 412, including volatile and non-volatile media, removable and non-removable media.

The system memory 428 may include computer system readable media in the form of volatile memory, such as random access memory (Random Access Memory, RAM) 430 and/or cache memory 432. Device 412 may include other removable/non-removable, volatile/nonvolatile computer system storage media. For example only, the storage system 434 may be configured to read and write non-removable, non-volatile magnetic media (not shown in FIG. 6 and is generally referred to as a "hard disk drive"). Although not shown in FIG. 6, a disk drive configured to read and write to a removable non-volatile disk (such as a "floppy disk"), and a removable non-volatile optical disk (such as a compact disc read-only memory) may be provided (Compact Disc Read-Only Memory, CD-ROM), digital video disc (Digital Video Disc-Read Only Memory, DVD-ROM) or other optical media) optical disc drive for reading and writing. In these cases, each drive may be connected to the bus 418 through one or more data media interfaces. The memory 428 may include at least one program product having a set of (eg, at least one) program modules configured to perform the functions of any embodiment of the present disclosure.

A program/utility tool 440 having a set of (at least one) program modules 442 may be stored in, for example, the memory 428. Such program modules 442 include but are not limited to an operating system, one or more application programs, other program modules, and program data Each of these examples or some combination may include the implementation of a network environment. The program module 442 generally performs the functions and/or methods in the embodiments described in the present disclosure.

The device 412 may also communicate with one or more external devices 414 (eg, keyboard, pointing device, display 424, etc.), and may also communicate with one or more devices that enable a user to interact with the device 412, and/or Device 412 can communicate with any device (eg, network card, modem, etc.) that can communicate with one or more other computing devices. Such communication may be performed through an input/output (I/O) interface 422. Moreover, the device 412 can also communicate with one or more networks (such as a local area network (Local Area Network, LAN), a wide area network (Wide Area Network, WAN), and/or a public network, such as the Internet) through the network adapter 420. As shown, the network adapter 420 communicates with other modules of the device 412 via the bus 418. It should be understood that although not shown in FIG. 6, other hardware and/or software modules may be used in conjunction with the device 412, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, disk arrays (Redundant Arrays) of Independent Drives (RAID) systems, tape drives and data backup storage systems, etc.

The processor 416 executes at least one functional application and data processing by running a program stored in the system memory 428, for example, to implement a prosthetic limb control method provided by an embodiment of the present disclosure, including: acquiring muscle signals at a prosthetic limb connection, Muscle signals include myoelectric signals and muscle impedance signals; identify the action type corresponding to the muscle signal; and generate control commands for the prosthesis according to the action type.

The processor 416 may also implement the technical solution applied to the prosthesis control method of the device provided by any embodiment of the present disclosure. The hardware structure and functions of the device can be explained in the content of the fourth embodiment.

Example 5

This embodiment also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the program is executed by a processor, a prosthetic limb control method as provided in an embodiment of the present disclosure is implemented. The method includes : Obtain the muscle signal at the joint of the prosthesis. The muscle signal includes the myoelectric signal and the muscle impedance signal; identify the action type corresponding to the muscle signal; and generate the control commands of the prosthesis according to the action type.

The computer storage medium of this embodiment may use any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above. Examples of computer-readable storage media (non-exhaustive list) include: electrical connections with one or more wires, portable computer disks, hard drives, RAM, read-only memory (Read-Only Memory, ROM), erasable Erasable Programmable Read-Only Memory (EPROM) or flash memory, optical fiber, portable CD-ROM, optical storage device, magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in conjunction with the instruction execution system, apparatus, or device.

The computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, and the computer-readable signal medium carries computer-readable program code. This propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, and the computer-readable medium may be sent, propagated, or transmitted for use by or in connection with an instruction execution system, apparatus, or device, A program used in conjunction with a device or device.

The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, radio frequency (Radio Frequency, RF), etc., or any suitable combination of the foregoing.

The computer program code for performing the operations of the present disclosure may be written in one or more programming languages or a combination of multiple programming languages, which includes object-oriented programming languages such as Java, Smalltalk, C++, It also includes conventional procedural programming languages-such as "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partly on the user's computer, as an independent software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the case where a remote computer is involved, the remote computer may be connected to the user's computer through any kind of network, including LAN or WAN, or may be connected to an external computer (eg, using an Internet service provider to connect through the Internet).

A computer-readable storage medium provided in this embodiment. Computer-executable instructions are not limited to the method operations described above, and can also perform related operations in the prosthetic limb control method of the device provided in any embodiment of the present disclosure. For the introduction of the storage medium, please refer to the content explanation in the fifth embodiment.

Claims

A prosthetic limb control method, including:

Acquiring muscle signals at the joint of the prosthesis, wherein the muscle signals include myoelectric signals and muscle impedance signals;

Identify the action type corresponding to the muscle signal;

The control instruction of the prosthesis is generated according to the action type.
The method according to claim 1, wherein the identifying the action type corresponding to the muscle signal comprises:

Superimposing the myoelectric signal and the muscle impedance signal to obtain a superimposed muscle signal;

The superimposed muscle signal is input into a pre-trained muscle action type recognition model to obtain the action type corresponding to the muscle signal.
The method according to claim 1, wherein the identifying the action type corresponding to the muscle signal comprises:

Extracting the characteristics of the EMG signal to obtain the EMG characteristic signal, and extracting the characteristics of the muscle impedance signal to obtain the muscle impedance characteristic signal;

Superimposing the EMG characteristic signal and the muscle impedance characteristic signal to obtain a superimposed characteristic signal;

The superimposed feature signal is input into a pre-trained feature action type recognition model to obtain the action type corresponding to the muscle signal.
The method according to claim 1, wherein the identifying the action type corresponding to the muscle signal comprises:

Input the EMG signal into a pre-trained EMG action type recognition model to obtain the EMG action type;

Input the muscle impedance signal into a pre-trained muscle impedance action type recognition model to obtain the muscle impedance action type;

The action type corresponding to the muscle signal is determined according to the myoelectric action type and the muscle impedance action type.
The method according to claim 2, wherein the muscle action type recognition model is trained in the following manner:

Acquiring the EMG sample signal, the muscle impedance sample signal, and the action type corresponding to the EMG sample signal and the muscle impedance sample signal;

Superimposing the EMG sample signal and the muscle impedance sample signal to obtain a superimposed muscle sample signal;

Taking the superimposed muscle sample signal as an input variable, the action type corresponding to the myoelectric sample signal and the muscle impedance sample signal as output variables, training a classifier model, and obtaining the muscle action type recognition model.
The method according to claim 3, wherein the characteristic action type recognition model is trained in the following manner:

Acquiring the EMG sample signal, the muscle impedance sample signal, and the action type corresponding to the EMG sample signal and the muscle impedance sample signal;

Extracting the characteristics of the EMG sample signal to obtain the EMG sample characteristic signal, and extracting the characteristics of the muscle impedance sample signal to obtain the muscle impedance sample characteristic signal;

Superimposing the EMG sample characteristic signal and the muscle impedance sample characteristic signal to obtain a superimposed muscle sample characteristic signal;

Taking the characteristic signal of the superimposed muscle sample as an input variable, the action type corresponding to the myoelectric sample signal and the muscle impedance sample signal as output variables, training a classifier model, and obtaining the characteristic action type recognition model.
The method according to claim 4, wherein the EMG type recognition model is trained by:

Acquiring the EMG sample signal and the action type corresponding to the EMG sample signal;

Using the EMG sample signal as an input variable and the action type corresponding to the EMG sample signal as an output variable, training a classifier model to obtain the EMG action type recognition model;

Train the muscle impedance action type recognition model by:

Acquiring a muscle impedance sample signal and an action type corresponding to the muscle impedance sample signal;

Using the muscle impedance sample signal as an input variable and the action type corresponding to the muscle impedance sample signal as an output variable, a classifier model is trained to obtain the muscle impedance action type recognition model.
The method according to any one of claims 1-7, wherein the EMG signal is obtained after processing according to the muscle impedance signal.
The method according to claim 8, wherein the EMG signal is obtained after processing the muscle impedance signal, including:

Acquiring the myoelectric signal to be screened, wherein the myoelectric signal to be screened and the muscle impedance signal are acquired synchronously;

If the muscle impedance signal is less than a preset threshold, the EMG signal to be screened is used as the EMG signal.
A prosthetic limb control device, including:

A muscle signal acquiring module, configured to acquire a muscle signal at the connection of the prosthesis, wherein the muscle signal includes an electromyographic signal and a muscle impedance signal;

An action type recognition module configured to recognize the action type corresponding to the muscle signal;

The control instruction generation module is configured to generate the control instruction of the prosthesis according to the action type.
A prosthetic limb control system includes an upper computer and a lower computer, the upper computer is provided with the artificial limb control device according to claim 10; the upper computer is in communication with the lower computer;

The lower computer is configured to collect muscle signals, the muscle signals include myoelectric signals and muscle impedance signals, and send the muscle signals to the upper computer.
The system according to claim 11, wherein the lower computer includes an acquisition module and a control module, and the control module is connected to the acquisition module and the upper computer respectively;

The control module is configured to control the collection module to synchronously collect the muscle signal and the muscle impedance signal, and send the myoelectric signal and the muscle impedance signal to the host computer.
The system of claim 12, wherein the acquisition module includes an electrode unit and an analog front end unit, the analog front end unit includes a modulator, a first channel, and a second channel; the electrode unit is configured to be placed on a prosthetic connection At the surface of the muscle tissue, the electrode unit is connected to the modulator, the first channel and the second channel respectively;

The modulator is configured to send the modulated signal to the electrode unit so that the electrode unit collects a mixed signal including the myoelectric signal and the muscle impedance signal, and the first channel is set to The mixed signal is processed to obtain the myoelectric signal, and the second channel is configured to process the mixed signal to obtain the muscle impedance signal.
A device, including:

At least one processor;

Memory, set to store at least one program;

When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1-9.
A computer-readable storage medium storing a computer program, which when executed by a processor implements the method according to any one of claims 1-9.