Hybrid dual-pathway upper limb prosthetic bionic control system and method based on fuzzy system

Abstract

The present invention provides a hybrid dual-pathway upper limb prosthetic bionic control system and method based on a fuzzy system. The system includes: a posture sensor that analyzes the user's environment and task status by acquiring their real-time body posture; an electroencephalogram (EEG) signal acquisition device that collects and analyzes the user's EEG signals; an electromyographic (EMG) sensor that complements the prosthetic's movements by detecting biosignals from the residual limb; a proprioception feedback module that provides the wearer with real-time proprioception feedback based on ambient temperature, contact pressure, and prosthetic angle status; an algorithmic decision module that performs fuzzy fusion analysis on signals transmitted by each device to achieve robust decision-making; and a motion control module that drives the prosthetic motor to perform corresponding movements using the signals output by the algorithmic decision module. This invention enhances the flexible and robust control of upper limb prostheses in complex multi-tasking scenarios, improving the stability and reliability of the entire system.

Description

Technical Field

[0001] The present invention relates to the technical field of bionic robot control, and in particular to a hybrid dual-pathway upper limb prosthetic bionic control system and method based on a fuzzy system. Background Art

[0002] With the development of science and technology, there are more and more electric prosthetic limb controls based on various signals, which can control the movement of prosthetic limbs by detecting and identifying the user's biological signals.

[0003] However, the current method of controlling prostheses based on a recognizable signal can make the prosthesis move, but it cannot drive the prosthesis quickly and efficiently. Moreover, due to the singleness of the control signal, the prosthesis can only recognize fewer movements, which to a certain extent cannot provide a good user experience.

[0004] Patent document CN109984875A (application number: CN201910359729.0) discloses a bionic mechanical prosthesis and control method, which relates to the field of prosthetic technology, including a prosthetic palm shell, a prosthetic finger mechanism and an external airbag mechanism; the motor control end inside the prosthetic palm shell is connected to a muscle electrical signal sensor via an electrical signal; a first pressure sensor and a thermistor are provided on the inner side of the finger segment structure; the first pressure sensor is connected to the external airbag mechanism via an electrical signal, and the thermistor is connected to a heating element via an electrical signal; by providing the thermistor and the heating element, the prosthesis can make the other party feel a comfortable temperature during driving etiquette movements, making it convenient to properly care for the other party during daily communication and avoid the other party holding a cold machine during the handshake; by providing an external airbag mechanism, the prosthesis can enable the user to sense the other party's strength during driving etiquette movements, thereby improving the user's sense of participation in use, and by providing the muscle electrical signal sensor, the prosthesis can be extended or tightened.

[0005] In the existing technology, the control method for upper limb prosthetic arms mainly adopts a single signal source or multiple signal sources for joint and simultaneous control. For example, a common control method is to control the upper limb prosthesis simultaneously through electromyographic signals and posture signals. This method improves the efficiency and accuracy of control to a certain extent. However, its main method is still to control the upper limb prosthesis through a single signal source, and does not train through multiple detectable signal sources for mixed control, resulting in low robustness and reliability of the signal source. Therefore, this technology still has a lot of room for improvement in control functions.

[0006] In order to improve the accuracy, speed and number of recognizable actions of prosthetic control, a new prosthetic control method needs to be proposed, which fuses multiple signal sources and trains them through algorithms to overcome some of the defects of current prosthetic control methods. Summary of the Invention

[0007] In view of the defects in the prior art, the purpose of the present invention is to provide a hybrid dual-pathway upper limb prosthetic bionic control system and method based on a fuzzy system.

[0008] The hybrid dual-pathway upper limb prosthetic bionic control system based on a fuzzy system provided by the present invention comprises:

[0009] Posture sensor: Analyzes the user's environment and task status by acquiring the user's real-time body posture;

[0010] EEG signal acquisition device: collects and analyzes the user's EEG signals and converts them into classifiable signals;

[0011] Myoelectric sensing device: supplements the movements performed by the prosthesis by detecting biological signals from the residual limb;

[0012] Proprioceptive feedback module: provides the wearer with real-time proprioceptive feedback based on external temperature, contact pressure and prosthetic angle status information;

[0013] Algorithm decision module: It performs fuzzy fusion analysis by receiving signals transmitted by various devices to achieve robust decision-making;

[0014] Motion control module: The signal output by the algorithm decision module drives the prosthetic motor to complete the corresponding action.

[0015] Preferably, the posture sensor is fixed on the instep of the user, and the posture sensor is driven by the feet to move to collect the acceleration, posture and direction information of the user during movement, and is connected to the signal processing device via Wi-Fi communication.

[0016] Preferably, the EEG signal acquisition device includes an EEG signal acquisition part and an EEG signal processing part. The EEG signal is collected by an EEG cap on the user, and then analyzed and processed by the EEG signal processing part through a filtering circuit and an A / D conversion circuit.

[0017] Preferably, the myoelectric sensing device obtains the myoelectric signal on the residual limb through an electromechanical sensor installed on the residual limb, transmits it to the signal processing device after enhanced rectification, and simultaneously stimulates the corresponding part of the residual limb through electrical stimulation by detecting the signal from the electronic skin.

[0018] Preferably, the proprioception feedback module includes:

[0019] Electronic skin tactile sensor: Receives external signals, including pressure and temperature, through the electronic skin covered on the prosthesis, and then transmits the signals to the signal processing device;

[0020] Electrical stimulation device: It provides real-time electrical stimulation feedback to the user by receiving the signal output by the signal processing device. By stimulating different positions of the residual limb and the size of the microcurrent, the electronic skin distributed in different positions responds to external signals.

[0021] According to the hybrid dual-pathway upper limb prosthetic bionic control method based on fuzzy system provided by the present invention, the following steps are performed:

[0022] Step 1: The posture sensor, EEG signal acquisition device, myoelectric sensor device, and proprioception feedback module collect signals from various parts of the user's body, and transmit the signals to the signal processing device of the algorithm decision module. Through multi-signal amplification and fuzzy system-based network training, a highly reliable, closed-loop control signal is fitted;

[0023] Step 2: The trained signal is transmitted to the upper limb prosthesis control device, which drives the servo motor to control the upper limb prosthesis through the signal, and the current state of the upper limb prosthesis and the signal from the electronic skin tactile sensor are returned to the signal processing device;

[0024] Step 3: The signal of the electronic skin tactile sensor is sent to the electrical stimulation device through the signal processing device of the algorithm decision module, and the change in the signal quantity of the electronic skin tactile sensor is fed back by stimulating the residual limb with electrical stimulation. At the same time, a closed loop is formed by the feedback signal and the current input signal to continuously control the upper limb prosthesis.

[0025] Preferably, the step 1 comprises:

[0026] Step 1.1: Use an EEG cap to collect EEG signals based on the position of the user's arm corresponding to the cerebral cortex, amplify and filter the collected EEG signals, perform feature extraction and motion pattern recognition on the preprocessed EEG signals, and output the processed signals;

[0027] Step 1.2: Install a posture sensor on the user's instep. The posture sensor collects 3D acceleration, 3D gyroscope, and 3D angle information of the healthy limb. The signals collected by the posture sensor are filtered and integrated to form the motion state of the healthy limb, and the processed signals are output.

[0028] Step 1.3: Install the electromyographic signal acquisition electrodes on the user's residual limb, perform noise elimination and classification on the acquired electromyographic signals, extract the corresponding electromyographic features and output them.

[0029] Preferably, the step 2 comprises:

[0030] Step 2.1: Distribute the flexible electrode material on various parts of the user's upper limb prosthetic palm;

[0031] Step 2.2: Identify and classify the collected pressure and temperature information;

[0032] Step 2.3: Output the corresponding processed signal;

[0033] Step 2.3: Based on the acquired signal, the electrical stimulation device controls the electrode to stimulate the user's residual limb.

[0034] Preferably, the step 3 comprises:

[0035] Step 3.1: Obtain input from multiple sources including EEG, posture, and EMG signals;

[0036] Step 3.2: Train and calculate the input signal through the neural network based on the fuzzy system;

[0037] Step 3.3: Control the prosthesis through the signals output by the neural network.

[0038] Preferably, the training of the input signal includes: fusing multi-source information using a fuzzy model, and calculating the control signals of the servo motors of the upper limb prosthesis in real time according to the states of the various input signals.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) The present invention combines EEG, myoelectricity, and posture signal sources for training, and forms a closed loop through output signal feedback, which can avoid errors caused by a single signal or multiple signals acting on the prosthesis. The training signal also has strong robustness, enabling the prosthesis to be used stably and naturally.

[0041] (2) The present invention forms a bidirectional conduction through electronic skin and EEG signal feedback, which makes it easier for users to send accurate signals based on the feedback signals during operation, thereby enhancing the user's experience of using the prosthesis. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0043] Figure 1 This is a system structure diagram of the intelligent upper limb prosthetic control system of the present invention;

[0044] Figure 2 It is a network structure diagram of the TS fuzzy model of the present invention;

[0045] Figure 3 This is the overall structural diagram of the intelligent upper limb prosthetic control system of the present invention;

[0046] Figure 4It is a training flow chart of the TS fuzzy model of the present invention;

[0047] FIG5 is a membership function of each input quantity of the present invention;

[0048] Figure 6 This is a flow chart of the upper limb prosthetic EEG signal acquisition and processing of the present invention;

[0049] Figure 7 This is a flow chart of the myoelectric signal acquisition and processing of the upper limb prosthesis of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0051] Example 1:

[0052] The hybrid dual-path upper limb prosthetic bionic control system based on a fuzzy system provided by the present invention includes: a posture sensor: which analyzes the user's environment and task status by acquiring the user's real-time body posture; an electroencephalogram (EEG) signal acquisition device: which collects and analyzes the user's EEG signals and converts them into classifiable signals; an electromyography (EMG) sensing device: which supplements the movements performed by the prosthesis by detecting biological signals on the residual limb; a proprioception feedback module: which provides the wearer with real-time proprioception feedback based on external temperature, contact pressure, and prosthetic angle status information; an algorithm decision module: which performs fuzzy fusion analysis on the signals transmitted by each device to achieve robust decision-making; and a motion control module: which drives the prosthetic motor to complete the corresponding movement through the signal output by the algorithm decision module.

[0053] The posture sensor is fixed to the user's instep. The foot-driven movement of the posture sensor collects acceleration, posture, and direction information during movement, and connects to a signal processing device via Wi-Fi. The EEG signal acquisition device comprises an EEG signal acquisition unit and an EEG signal processing unit. EEG signals are collected via an EEG cap worn by the user, then filtered and processed by the EEG signal processing unit. The myoelectric sensing device uses bioelectric sensors mounted on the residual limb to capture myoelectric signals from the residual limb. After enhanced and rectified, it is transmitted to the signal processing unit. Simultaneously, the device detects signals from the electronic skin and electrically stimulates the corresponding areas of the residual limb. The proprioceptive feedback module comprises an electronic skin tactile sensor that receives external signals, including pressure and temperature, through the electronic skin covering the prosthetic limb and transmits the signals to the signal processing unit. An electrical stimulation unit receives signals from the signal processing unit and provides real-time electrical stimulation feedback to the user. The stimulation of different locations on the residual limb and the magnitude of the microcurrent correspond to the response of the electronic skin to external signals.

[0054] According to the hybrid dual-path upper limb prosthesis bionic control method based on fuzzy system provided by the present invention, the following steps are performed: Step 1: collect signals from various parts of the user through a posture sensor, an EEG signal acquisition device, an electromyography sensor device and a proprioception feedback module, and transmit the signals to the signal processing device of the algorithm decision module, and fit a highly reliable, closed-loop control signal through multi-signal amplification and fuzzy system-based network training; Step 2: transmit the trained signal to the upper limb prosthesis control device, control the upper limb prosthesis through the signal-driven servo motor, and return the current state of the upper limb prosthesis and the signal of the electronic skin tactile sensor to the signal processing device; Step 3: send the signal of the electronic skin tactile sensor to the electrical stimulation device through the signal processing device of the algorithm decision module, and feedback the change in the signal amount of the electronic skin tactile sensor by stimulating the residual limb with electrical stimulation, and at the same time continuously control the upper limb prosthesis by forming a closed loop through the feedback signal and the current input signal.

[0055] The step 1 includes: step 1.1: using an EEG cap to collect EEG signals according to the position of the user's arm corresponding to the cerebral cortex, amplifying and filtering the collected EEG signals for pre-processing, performing feature extraction and motion pattern recognition processing on the pre-processed EEG signals, and outputting the processed signals; step 1.2: installing a posture sensor on the user's instep, collecting three-dimensional acceleration, three-dimensional gyroscope, and three-dimensional angle information of the healthy limb through the posture sensor, screening and integrating the signals collected by the posture sensor to form the movement state of the healthy limb, and outputting the processed signals; step 1.3: installing an electromyographic signal acquisition electrode on the user's residual limb, performing noise elimination and classification on the acquired electromyographic signals, extracting corresponding electromyographic features and outputting them.

[0056] The step 2 includes: step 2.1: distributing the flexible electrode material on various parts of the palm of the user's upper limb prosthesis; step 2.2: identifying and classifying the collected pressure and temperature information; step 2.3: outputting the corresponding processed signal; step 2.3: controlling the electrode piece to stimulate the user's residual limb through the electrical stimulation device according to the acquired signal.

[0057] Step 3 includes: Step 3.1: Acquiring multi-source signals, including EEG, posture, and myoelectric signals; Step 3.2: Training and calculating the input signals using a fuzzy-based neural network; and Step 3.3: Controlling the prosthesis using the neural network's output signals. Input signal training involves fusing multi-source information using a fuzzy model and, based on the states of each input signal, calculating control signals for each servo motor in the upper limb prosthesis in real time.

[0058] Example 2:

[0059] Example 2 is a preferred example of Example 1.

[0060] The present invention provides a hybrid dual-path intelligent upper limb prosthetic bionic control system and method based on fuzzy system, referring to Figure 1 and Figure 2 The method shown includes steps of recognizing EEG, EMG and posture signals as well as signal processing and training steps.

[0061] like Figure 3 The attitude sensor is a device that measures the three-dimensional acceleration, three-dimensional attitude angle and angular velocity of an object. The main components of the attitude sensor include motion sensors such as a three-axis gyroscope, a three-axis accelerometer, and a three-axis electronic compass. The three-axis gyroscope simultaneously measures the position, movement trajectory and acceleration in six directions, while the three-axis accelerometer determines the motion state of the object by measuring spatial acceleration. The three-axis electronic compass can detect the inclination and direction of the object through a magnetoresistive sensor and a two-axis inclination sensor, and then obtain the object's three-dimensional attitude and orientation data through the embedded processor.

[0062] Fuzzy neural networks are the product of the integration of fuzzy theory and neural networks. They combine the advantages of both theories, integrating learning, association, recognition, and information processing. A fuzzy neural network is a neural network with fuzzy weights or fuzzy input signals. Fuzzy neural networks are generally divided into four layers: the input layer, the fuzzification layer, the fuzzy inference layer, and the defuzzification layer. Each node and all parameters in the network have distinct physical meanings, so the initial values ​​of these parameters can be determined based on system or qualitative knowledge. The aforementioned learning algorithm can then quickly converge to the desired input-output relationship.

[0063] like Figure 4 The present invention primarily measures and acquires the user's electromyography, electroencephalography, and posture signals, then uses network training to judge and fit each signal emitted by the user. This results in a more reliable signal after training, allowing the prosthesis to more accurately reach the user's desired position during movement. Furthermore, by distributing corresponding electronic skin to the prosthetic upper limb, the prosthetic upper limb is equipped with the ability to sense signals such as pressure and temperature, and achieves two-way communication of signals through electrical stimulation.

[0064] like Figure 6 As shown, the user wears an EEG cap. When collecting information, the multiple sensor probes on the EEG cap obtain EEG signals from different areas of the brain and perform denoising on these signals. A combination of low-pass filtering and high-pass filtering is used to remove biological signal noise and environmental noise contained in the brain waves. At the same time, the eye point filter also needs to be removed, and then converted through an A / D converter and sent to the EEG signal processing part.

[0065] The denoised EEG signal is divided into multiple segments for each task using a sliding window technique for analysis. The data within the window is first subjected to Laplace filtering and an 8-30Hz bandpass filter. Then, a runaway feature extraction method combining wavelet decomposition and common spatial patterns and a long-short-term memory network are used for pattern recognition. Various classification results are then weighted averaged to obtain the final recognition pattern for the wearer's command, which is then output as a signal. The reference potential of the EEG signal can be selected from signals in the prefrontal, frontal, and parietal lobes.

[0066] like Figure 7Myoelectric signal recognition can collect surface electromyographic signals during independent flexion and extension, internal and external rotation, and elbow flexion and extension of the user's healthy limb, as well as during independent hand opening and closing movements. Electromyographic electrodes can be placed on eight myoelectric surface areas associated with hand opening and closing, wrist flexion and extension, wrist internal and external rotation, and elbow flexion and extension: the brachialis, biceps brachii, brachioradialis, extensor carpi radialis longus, flexor carpi ulnaris, and extensor carpi ulnaris. The electrodes collect eight channels of myoelectric signals. The surface electromyographic signals collected during the training phase of independent joint movement are baseline-removed, full-wave rectified, and low-pass filtered. The pre-processed surface electromyographic signals are integrated and calculated as the muscle activation matrix f(E) to create a muscle synergy model. By creating the muscle synergy model and utilizing support vector regression, a coordinated activation model of the hand, wrist, and elbow joint activation coefficient sequences and joint angle information is established. This achieves simultaneous estimation of the hand, wrist, and elbow joint angles of the upper limb, simulating the training and estimation phases.

[0067] During the simulation training phase, a surface electromyography (EMG) training dataset and corresponding joint angles are collected. A 3D motion capture system is then used to simultaneously capture the corresponding motion angles of the hand, wrist, and elbow joints. These motion angles are then subjected to a sliding smoothing filter to calculate a synergy matrix. An activation coefficient matrix is ​​then calculated from the EMG training dataset generated by the independent motion of each degree of freedom. The activation coefficients are normalized using the maximum value of each channel and used as input to a vector regression algorithm. The angles corresponding to the four degrees of freedom (DOF) of the hand, wrist, and elbow are used as target values ​​for training. Specifically, the activation coefficient matrix is ​​first low-pass filtered and normalized to the [0, 1] interval. After preprocessing, it is used as input to the synergistic activation model. The measured angles of the healthy limb, obtained using the 3D motion capture system, are also normalized to the same interval and used as target values ​​to complete the model training.

[0068] During the estimation stage, the activation coefficient is extracted from the collected electromyographic signal, normalized, and input into the vector regression algorithm. The output is used as the estimated angle value to control the movement of the bionic hand. The estimated angle value is subjected to sliding average filtering. If the angle estimate exceeds the maximum value of the actual joint angle, the angle at this time is set to the maximum value of the actual joint angle.

[0069] During the motion estimation phase, the surface electromyography (EMG) signals are collected in real time. The muscle activation matrix f(E) is extracted using the aforementioned method. Combined with the cooperating element matrix obtained during the model training phase, the activation coefficient sequence is extracted. Similar to the training phase, preprocessing and normalization are performed before inputting the sequence into the trained SVR co-activation model. Finally, the output values ​​are denormalized using the measured angle normalization parameters from the training phase to obtain the EMG control signal.

[0070] The activation function is then extracted from the collected electromyographic signals, normalized, and input into the model. The output is then transmitted as a signal to the signal processing device. By placing posture sensors on the user's insteps and hands, the user's movement state and patterns are determined by collecting posture signals, including the direction and speed of the feet. A posture sensor can also be placed on the back of the hand to coordinate bimanual operations. By acquiring multiple signals, a fuzzy neural network model can be used to fuse multi-source information. The control test of the prosthetic arm is summarized as a fuzzy rule table. Each rule in the fuzzy rule table is then extracted and converted into input and output samples to form a sample set. A three-layer neural network model is initialized and trained offline using the sample set. The trained neural network model is then defined as a fuzzy neural network planning model.

[0071] The fuzzy rule model describes the mapping relationship between input state quantities and output state quantities. When the control device receives a control signal to move, the control signal is mainly based on EEG signals, posture signals, and myoelectric signals. The control signal is improved and adjusted through neural network training. Define the EEG signal quantity as d, the posture signal quantity as p, and the myoelectric signal quantity as t. The input state quantities are fuzzified into the linguistic variable sets D, P, and T respectively:

[0072] D={D1,D2,…D i ,…,D l}

[0073] P={P1,P2,…P j ,…,P m}

[0074] T={T1,T2,…T k ,…,T n}

[0075] Among them, l, m, and n are the number of language variables of EEG signal quantity, posture signal quantity, and EMG signal quantity, respectively.

[0076] Define the output quantity as a and fuzzify the output quantity into a set of language variables A:

[0077] A={A1,A2,…A i ,…,A k}

[0078] Where k is the number of output language variables.

[0079] In order to facilitate the subsequent unified analysis, the language variable D l 、P m 、T n and A kInstead of conventional language variables, such as negative large NB, negative small NS, positive small PS, positive large PB, etc. According to fuzzy system theory, firstly, a membership function is established for each input variable and output variable, such as Figure 5a 、 Figure 5b 、 Figure 5c As shown. And combined with the control rules to establish a fuzzy rule table. Here the membership function uses the Gaussian membership function:

[0080]

[0081] Where x is the input variable, σ is the standard deviation of the normal distribution, and c is the mean μ in the normal distribution.

[0082] According to the established fuzzy rule table and the membership function of the variables, a control signal planning model can be established, which describes the mapping relationship from the input state quantity to the output state quantity, that is, a ijk =f(d i ,p j ,t k ), where d i 、p j , t k 、a ijk are the maximum membership of each language variable respectively. The samples corresponding to each rule in the fuzzy rule table constitute the set of input and output.

[0083] A three-layer neural network model was established, in which the number of neurons in the input layer equals the number of input states, the number of neurons in the hidden layer is N, and the number of neurons in the output layer equals the number of output states. The weight vectors from the input layer to the hidden layer and from the hidden layer to the output layer are w1 and w2, respectively, and the bias vectors for the hidden layer and output layer are b1 and b2, respectively. The weight and bias vectors of the neural network were trained using a backpropagation algorithm. Finally, the trained neural network model was used as a fuzzy neural network control signal planning model. When the upper limb prosthesis control device receives the fuzzy network output signal, it can move the upper limb prosthesis along a predetermined trajectory based on the control signal, thereby controlling the movement of the bionic hand. Simultaneously, the estimated angle is subjected to a sliding average filter. If the angle estimate exceeds the maximum actual joint angle, the angle at that time is set to the maximum actual joint angle. The control device then controls the upper limb prosthesis to perform the corresponding movement. The electronic skin covering the upper limb prosthesis senses the changes in pressure and temperature of the electronic skin during the process of the upper limb prosthesis completing the corresponding action. The electronic skin can be distributed on the fingertips, inner sides of the fingers, inner sides of the palm and other parts of the upper limb prosthesis. The pressure, temperature and other signals are detected by the sensors of the electronic skin and transmitted back through electrical stimulation. The signals are classified by setting the frequency of the electrical stimulation, and the corresponding positions of the residual limb are stimulated accordingly.

[0084] Those skilled in the art will appreciate that, in addition to implementing the system, device, and various modules provided by the present invention in purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, and the like by logically programming the method steps. Therefore, the system, device, and various modules provided by the present invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; the modules for implementing various functions can also be considered both software programs for implementing the method and structures within the hardware component.

[0085] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A hybrid dual-pathway upper limb prosthetic bionic control system based on fuzzy system, characterized in that: include: Posture sensor: Analyzes the user's environment and task status by acquiring the user's real-time body posture; EEG signal acquisition device: collects and analyzes the user's EEG signals and converts them into classifiable signals; Myoelectric sensing device: supplements the movements performed by the prosthesis by detecting biological signals from the residual limb; Proprioceptive feedback module: provides the wearer with real-time proprioceptive feedback based on external temperature, contact pressure and prosthetic angle status information; Algorithm decision module: It performs fuzzy fusion analysis by receiving signals transmitted by various devices to achieve robust decision-making; Motion control module: drives the prosthetic motor to complete the corresponding action through the signal output by the algorithm decision module; Initialize a three-layer neural network model, use the sample set to train the neural network offline, obtain the offline trained neural network model, and define it as a fuzzy neural network planning model; The fuzzy rule model describes the mapping relationship between input state quantities and output state quantities. When the control device receives a control signal to move, the control signal is mainly improved and adjusted by training the neural network based on the EEG signal, posture signal, and myoelectric signal. The EEG signal quantity is defined as d, the posture signal quantity as p, and the myoelectric signal quantity as t. The input state quantities are fuzzified into the language variable sets D, P, and T respectively: D={D1,D2,…D i ,…,D l } P={P1,P2,…P j ,…,P m } T={T1,T2,…T k ,…,T n } Among them, l, m, and n are the number of language variables of EEG signal quantity, posture signal quantity, and EMG signal quantity respectively; Define the output quantity as a and fuzzify the output quantity into a set of language variables A: A={A1,A2,…A i ,…,A k } Where k is the number of output language variables; Use language variable D l 、P m 、T n and A k Instead of conventional linguistic variables, according to fuzzy system theory, a membership function is first established for each input variable and output variable, and a fuzzy rule table is established in combination with the control rules; the membership function used here is a Gaussian membership function: Where x is the input variable, σ is the standard deviation of the normal distribution, and c is the mean μ in the normal distribution; A three-layer neural network model was established, in which the number of neurons in the input layer was the number of input state variables, the number of neurons in the hidden layer was N, and the number of neurons in the output layer was the number of output variables. The weight vectors from the input layer to the hidden layer and from the hidden layer to the output layer were w1 and w2, respectively, and the bias vectors of the hidden layer and the output layer were b1 and b2, respectively. The weight vectors and bias vectors of the neural network were trained using the BP algorithm. Finally, the trained neural network model was used as a fuzzy neural network control signal planning model. When the upper limb prosthesis control device received the fuzzy network output signal, it could move the upper limb prosthesis along a predetermined trajectory according to the control signal, thereby controlling the movement of the bionic hand. At the same time, the estimated angle is subjected to sliding average filtering. If the angle estimation value exceeds the maximum value of the actual joint angle, the angle at this time is set to the maximum value of the actual joint angle; the control device controls the upper limb prosthesis to complete the corresponding action, and senses the changes in pressure and temperature of the electronic skin covered on the upper limb prosthesis during the process of the upper limb prosthesis completing the corresponding action. The electronic skin can be distributed on the fingertips, inner sides of the fingers, and inner sides of the palm of the upper limb prosthesis. The pressure and temperature signals are detected by the sensors of the electronic skin and transmitted back through electrical stimulation. The signals are classified by setting the frequency of the electrical stimulation, and the corresponding positions of the residual limb are stimulated accordingly.

2. The hybrid dual-path upper limb prosthetic bionic control system based on fuzzy system according to claim 1, characterized in that: The posture sensor is fixed on the instep of the user, and the feet drive the posture sensor to move to collect the acceleration, posture and direction information of the user during movement, and connect to the signal processing device through WIFI communication.

3. The hybrid dual-path upper limb prosthetic bionic control system based on fuzzy system according to claim 1, characterized in that: The EEG signal acquisition device includes an EEG signal acquisition part and an EEG signal processing part. The EEG signal is collected by an EEG cap on the user, and then passes through a filtering circuit and an A / D conversion circuit and is analyzed and processed by the EEG signal processing part.

4. The hybrid dual-path upper limb prosthetic bionic control system based on fuzzy system according to claim 1, characterized in that: The myoelectric sensing device obtains the myoelectric signals on the residual limb through the bioelectric sensor installed on the residual limb, transmits them to the signal processing device after enhancement and rectification, and stimulates the corresponding part of the residual limb through electrical stimulation by detecting the signals from the electronic skin.

5. The hybrid dual-path upper limb prosthetic bionic control system based on fuzzy system according to claim 1, characterized in that: The proprioception feedback module includes: Electronic skin tactile sensor: Receives external signals, including pressure and temperature, through the electronic skin covered on the prosthesis, and then transmits the signals to the signal processing device; Electrical stimulation device: It provides real-time electrical stimulation feedback to the user by receiving the signal output by the signal processing device. By stimulating different positions of the residual limb and the size of the microcurrent, the electronic skin distributed in different positions responds to external signals.

Images