Rehabilitation manipulator based on myoelectricity feedback and electrical stimulation self-adaptive control method

Abstract

The invention discloses a rehabilitation manipulator based on myoelectricity feedback and an electrical stimulation self-adaptive control method. The method comprises the following steps: S1, constructing a myoelectricity acquisition-rehabilitation manipulator-neuromuscular electrical stimulation combined rehabilitation system; s2, performing data preprocessing on the surface electromyogram signals; s3, performing loop iteration on the long and short term memory network parameters by using a frost ice optimization algorithm, and constructing a long and short term memory network model of the optimal parameters; s4, predicting the surface electromyogram signals by using the constructed long and short-term memory network model; s5, constructing a three-dimensional myoelectricity adaptive control system according to the prediction signal; s6, generating a fuzzy control rule based on the adaptive fuzzy control system; and S7, controlling the combined rehabilitation system based on the control strategy. According to the rehabilitation manipulator based on myoelectricity feedback and the electrical stimulation self-adaptive control method, secondary injury caused by non-timeliness of subjective parameter regulation and control in the existing rehabilitation process can be solved, and the rehabilitation efficiency and the rehabilitation effect can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of adaptive control of rehabilitation medical robots, and in particular to a rehabilitation manipulator based on electromyographic feedback and an adaptive control method of electrical stimulation. Background Art

[0002] At present, traditional treatment methods mainly use physical rehabilitation therapy, and the most common one is exercise therapy. This treatment method requires the therapist to manually pull the patient's affected limb with the assistance of rehabilitation equipment and rehabilitation aids to assist the patient in completing the movement task. Repeated assisted exercise can help patients avoid the disuse and degeneration of motor organs caused by motor dysfunction, and at the same time it is very helpful for the recovery of their motor function.

[0003] This therapy has certain defects. It requires a large number of professionally trained therapists with rich clinical experience, which not only increases the difficulty of training clinical therapists, but also causes therapists to face great work pressure. This traditional treatment method cannot meet the growing number of stroke patients and the needs of post-stroke rehabilitation treatment. Therefore, it is necessary to explore reliable and intelligent treatment methods to alleviate post-stroke motor dysfunction.

[0004] In response to the damaged neural pathways caused by stroke, functional electrical stimulation (FES) is applied to the post-operative movement disorder rehabilitation training process. By applying continuous electrical stimulation to the muscles innervated by the patient's damaged nerves, the muscles are contracted and limb movements are generated, thereby restoring the patient's limb motor function. At the same time, sensory feedback stimulates the brain's motor relearning and promotes the repair of the entire damaged motor neural pathway.

[0005] For limb movement, the emergence of wearable rehabilitation mechanical exoskeletons has led to the development of passive limb rehabilitation. In the field of hand function rehabilitation, wearable rehabilitation manipulators are mainly driven by pneumatic, rope drive and motor push rod to drive the patient's affected hand for rehabilitation training.

[0006] Both of the above two rehabilitation methods can achieve relatively good hand function rehabilitation training effects. However, combining rehabilitation robots with functional electrical stimulation can well integrate the advantages of both and solve the problems of inaccurate open-loop control of electrical stimulation and low activation of specific muscles in rehabilitation manipulators.

[0007] During the whole rehabilitation cycle, the muscle state continues to change, so therapists with rich clinical experience are needed to monitor the patient's rehabilitation process in real time. When the patient's muscle state is not suitable for the current rehabilitation mode, necessary adjustments need to be made to prevent secondary injuries. This leads to a significant increase in the workload of therapists. Predicting the muscle state in advance and adjusting the control strategy is the starting point for the combination of the two.

[0008] Single electrical stimulation rehabilitation is often prone to muscle fatigue, and because there is no rigid structure to guide the extension of the finger movement, it is easy to cause damage. Single robot control cannot perfectly form a muscle-limb end control loop. Compensating FES-driven muscles through a manipulator can reduce the FES drive intensity, while FES stimulation can provide better muscle activation. The combination of the two provides a better treatment plan for post-stroke rehabilitation. Summary of the invention

[0009] The purpose of the present invention is to provide a rehabilitation manipulator based on electromyographic feedback and an electric stimulation adaptive control method, which can ensure the safety of the rehabilitation process, reduce the possibility of muscle damage, help improve the rehabilitation training effect, and accelerate the recovery of neuroplasticity function and the rehabilitation process.

[0010] To achieve the above object, the present invention provides a rehabilitation manipulator based on electromyographic feedback and an electric stimulation adaptive control method, comprising the following steps:

[0011] S1. Construct a myoelectric acquisition-rehabilitation manipulator-neuromuscular electrical stimulation combined rehabilitation system;

[0012] S2, using complete set empirical mode decomposition and adaptive noise algorithm to preprocess the surface electromyography signal;

[0013] S3, using the frost optimization algorithm to cyclically iterate the long short-term memory network parameters and construct a long short-term memory network model with optimal parameters;

[0014] S4, using the constructed long short-term memory network model to predict each decomposition mode of the surface electromyography signal;

[0015] S5, constructing a three-dimensional myoelectric adaptive control system based on the predicted signal as feedback;

[0016] S6, Generate fuzzy control rules based on adaptive fuzzy control system and electromyographic feedback;

[0017] S7. Control the combined rehabilitation system based on the control strategy.

[0018] Preferably, in S1, electromyographic signals are collected by electromyographic acquisition equipment, and the collected electromyographic signals are uploaded to the host computer for processing and analysis; RS485 one-to-two serial port communication is used to establish communication between the host computer and the rehabilitation manipulator and functional electrical stimulation, and the rehabilitation manipulator is used to provide passive rehabilitation of the hand; and active activation is provided through functional electrical stimulation.

[0019] Preferably, in S2, DC component detection is used to remove the DC component, an adaptive 50 Hz notch filter is used to remove power frequency interference, fourth-order Butterworth high- and low-pass filters are used to remove high and low-frequency noise in the signal, and the processed surface electromyography signal is decomposed into modal components of different frequency bands using a complete set empirical mode decomposition and an adaptive noise model.

[0020] Preferably, the processed surface electromyographic signal is decomposed from high to low using complete set empirical mode decomposition and adaptive noise model to obtain the corresponding intrinsic mode component IMF 1 、IMF 2 ...IMF n ; Decompose the one-dimensional surface electromyography signal data into multi-dimensional data in different frequency ranges so that the eigenvalues ​​fall within different frequency ranges. During the decomposition process, the first-order IMF component is calculated on an overall average basis to obtain the final first-order IMF component, and then the above operation is repeated on the residual part.

[0021] Preferably, in S3, the intrinsic mode components of each surface electromyographic signal are reorganized and normalized to form a long short-term memory network model input data set;

[0022] The initialized long short-term memory network model parameters constitute random frost particles, and the frost particles are iteratively optimized according to the particle update calculation formula to detect whether the particles are in the optimal position. If they are in the optimal position, the output is the optimal parameter, otherwise iterative optimization continues. The particle update calculation formula is:

[0023]

[0024] Preferably, in S4, the optimized long short-term memory network model is constructed using the optimal parameters, and each modal component after reorganization and normalization is predicted at a time of Δt to obtain the modal component IMF after a time of Δt. 1 (T+Δt), IMF 2 (T+Δt)……IMF 3 (T+Δt);

[0025] For each modal component IMF 1 (T+Δt), IMF 2 (T+Δt)……IMF 3(T+Δt) is superimposed according to the coefficient to obtain the predicted surface electromyographic signal.

[0026] Preferably, in S5, feature extraction is performed on the predicted surface electromyographic signal and a three-dimensional model based on the predicted surface electromyographic signal as feedback is constructed; the median frequency MF, the mean power frequency MPF, the integrated electromyographic value iEMG, and the electromyographic root mean square value RMS are extracted respectively,

[0027]

[0028] The three-dimensional model (X, Y, Z) is constructed by the following formula:

[0029]

[0030] In the formula, X f is the fatigue degree change in the frequency domain, X Am is the change of fatigue degree in time domain, n is the number of modes, Y is the degree of muscle activation, Z is the muscle activation state, iEMGA is the integrated EMG value of the agonist muscle under the corresponding action, and iEMGB is the integrated EMG value of the antagonist muscle under the corresponding action; the system adaptively adjusts parameter changes according to the feedback formed by the predicted muscle state.

[0031] Preferably, in S6, the predicted three-dimensional model is used as the input of the fuzzy rule: when the predicted muscle fatigue degree X increases, the movement frequency of the rehabilitation manipulator decreases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all decrease; when the predicted muscle fatigue degree X decreases, the movement frequency of the rehabilitation manipulator increases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all increase; when the predicted muscle activation degree Y increases, the movement frequency of the rehabilitation manipulator increases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all decrease; when the predicted muscle activation degree Y decreases, the movement frequency of the rehabilitation manipulator decreases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all increase; when the predicted activation state Z is normal, the joint rehabilitation system continues to work; when the predicted activation state Z is abnormal, the joint rehabilitation system stops working.

[0032] Preferably, in S7, the three-dimensional input (X, Y, Z) of the predicted state is fuzzified by a fuzzy controller, and fuzzy decision is made according to the fuzzy control table constructed by three-dimensional electromyography fuzzy rules. The required output result is obtained after fuzzy reasoning, and it is defuzzified. According to the adaptive rules, it is determined whether the current muscle state matches the rehabilitation system parameters, and the system is adaptively adjusted to finally obtain parameters (F, U, I, H), where F represents the movement frequency of the rehabilitation manipulator, U represents the electrical stimulation voltage, I represents the electrical stimulation current, and H represents the electrical stimulation frequency; the parameters of the combined rehabilitation system are adjusted to match the subsequent muscle state.

[0033] Therefore, the beneficial effects of the present invention using the above-mentioned rehabilitation manipulator based on electromyographic feedback and the electric stimulation adaptive control method are as follows:

[0034] (1) The combined rehabilitation system of the present invention fully combines the advantages of passive rehabilitation of rehabilitation robots and active activation of nerves under active rehabilitation with functional electrical stimulation.

[0035] (2) In the present invention, CEEMDAN is used to decompose the electromyographic signal to effectively extract the frequency band characteristics of the signal; the RIME algorithm is used to optimize the parameters of the LSTM model to improve the model prediction ability; and the surface electromyographic signal is jointly predicted in time and frequency by combining CEEMDAN-RIME-LSTM to improve the prediction accuracy of the electromyographic signal and provide effective muscle state for system adaptation.

[0036] (3) The present invention uses adaptive fuzzy control as the control framework of the combined rehabilitation system, fully considering the time-varying and complexity of the muscle state, and making the control process more natural and smooth based on sufficient fuzzy rules, thereby solving the problems of simple training and unsatisfactory rehabilitation effects in existing hand rehabilitation strategies and systems.

[0037] (4) The present invention constructs three-dimensional electromyographic fuzzy rules, fully considers the three muscle states in the rehabilitation process, and makes real-time adjustments based on fatigue level, activation level, and activation state. This allows the rehabilitation system to adaptively adjust parameters based on the time-varying nature of the patient's muscles during rehabilitation, making the entire joint rehabilitation system individually adaptable and significantly improving system safety.

[0038] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a step diagram of a rehabilitation manipulator based on electromyographic feedback and an electric stimulation adaptive control method of the present invention;

[0040] Figure 2 It is a structural block diagram of the myoelectric acquisition-rehabilitation manipulator-neuromuscular electrical stimulation combined rehabilitation system of an embodiment of the present invention;

[0041] Figure 3 It is a collaborative control framework of the myoelectric acquisition-rehabilitation manipulator-neuromuscular electrical stimulation combined rehabilitation system provided by an embodiment of the present invention;

[0042] Figure 4 is a flow chart of preprocessing of surface electromyography signals provided by an embodiment of the present invention;

[0043] Figure 5 It is a flow chart of electromyography prediction based on CEEMDAN-RIME-LSTM provided by an embodiment of the present invention;

[0044] Figure 6 It is a block diagram of a three-dimensional myoelectric adaptive control system constructed according to a prediction signal provided by an embodiment of the present invention;

[0045] Figure 7 It is a three-dimensional myoelectric fuzzy rule block diagram constructed based on the three-dimensional myoelectric adaptive fuzzy control system of an embodiment of the present invention;

[0046] Figure 8 It is a block diagram of controlling the combined rehabilitation system based on the control strategy according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0048] Embodiment 1

[0049] like Figure 1 As shown, the present invention provides a rehabilitation manipulator based on electromyographic feedback and an electric stimulation adaptive control method, comprising the following steps:

[0050] S1. Construct an electromyographic acquisition-rehabilitation manipulator-neuromuscular electrical stimulation combined rehabilitation system.

[0051] like Figure 2 As shown in the figure, the system includes the device side, data side, control side and electromyographic feedback. The device side is a four-channel functional electrical stimulation, a rehabilitation manipulator, and a Delsys electromyographic acquisition device. The data side is divided into two parts: surface electromyographic signal preprocessing and surface electromyographic signal prediction. The control side uses fuzzy adaptive control.

[0052] The combined rehabilitation system constructs electromyographic feedback according to the muscle change state of the rehabilitation patient, avoiding secondary injuries caused by the mismatch between the current muscle state and the rehabilitation equipment parameters during the rehabilitation process, ensuring the safety of the rehabilitation process while improving the rehabilitation efficiency.

[0053] like Figure 3 As shown, in this embodiment, the Delsys Trigno Wireless surface electromyography acquisition device is used and pasted on the muscle position required for rehabilitation movements. Functional electrical stimulation patches are pasted on both sides in a cross-shaped manner to reduce the interference of electrical stimulation on the electromyographic signal while producing good stimulation to the muscles. A rehabilitation manipulator is worn on the affected hand. The collected surface electromyographic signal is uploaded to the host computer using Bluetooth for subsequent processing. The processed surface electromyographic signal is predicted, and the predicted signal is used as the input of the fuzzy adaptive control system to obtain the required system parameters, and the system parameters are communicated in a one-to-two joint manner using RS485.

[0054] S2. Use complete set empirical mode decomposition and adaptive noise algorithm to preprocess the surface electromyography signal.

[0055] like Figure 4 As shown in the figure, the surface electromyography signal collected from the affected hand is processed to remove the DC component to reduce the risk of signal amplitude and zero-frequency part roll-off; an adaptive 50Hz notch filter is used to eliminate the 50Hz power frequency interference in the environment to further improve the accuracy of the signal; finally, a fourth-order Butterworth filter is used to perform high-pass and low-pass filtering on the original surface electromyography signal to filter out high-frequency and low-frequency noise signals, thereby improving and retaining useful frequency band signals.

[0056] The preprocessed signal is decomposed using CEEMDAN. CEEMDAN decomposition obtains the final first-order IMF component by performing an overall average calculation on the first-order IMF component, and then repeats the above operation on the residual part, which effectively solves the problem of white noise transfer from high frequency to low frequency.

[0057] The relevant calculation formula is:

[0058] E i (·) is the i-th eigenmode component obtained after EMD decomposition, and the i-th eigenmode component obtained by CEEMDAN decomposition is v j is a Gaussian white noise signal that satisfies the standard normal distribution, j=1, 2, ..., N is the number of times white noise is added, ε is the standard table of white noise, and y(t) is the signal to be decomposed. Adding Gaussian white noise to the signal to be decomposed y(t) yields the signal y(t)+(-1) q εv j (t), where q = 1, 2, ... Perform EMD decomposition on the new signal to obtain the first-order eigenmode component C 1 .

[0059] The first intrinsic mode component of CEEMDAN decomposition is obtained by averaging the N generated mode components.

[0060] The formula is:

[0061] Compute the residual after removing the first modal component:

[0062] In r 1 (t) is added with positive and negative paired Gaussian white noise to obtain a new signal, and the new signal is used as the carrier for EMD decomposition to obtain the first-order modal component D 1 , from which the second eigenmode component of CEEMDAN decomposition can be obtained:

[0063] Compute the residual after removing the second modal component:

[0064] Repeat the above steps until the residual signal is a monotonic function and cannot be further decomposed, and the algorithm ends. At this time, the number of intrinsic mode components obtained is K, and the original signal is decomposed into:

[0065] After CEEMDAN decomposition, the corresponding intrinsic mode component IMF is obtained 1 、IMF 2 ...IMF n . Decompose the one-dimensional data into multi-dimensional data in different frequency ranges so that the eigenvalues ​​fall within different frequency ranges.

[0066] The decomposed signal is subjected to feature extraction, and characteristic values ​​such as power spectral density, integrated electromyographic value, root mean square value, average power frequency, and median frequency are extracted for the modal components in different frequency ranges.

[0067] The integrated EMG is the sum of the areas under the curve per unit time after the EMG signal is rectified and filtered. It can reflect the changes in the strength of the EMG signal over time. The integrated EMG value can be used to analyze the contraction characteristics of the muscle per unit time.

[0068] The formula is:

[0069] The root mean square value (RMS) refers to the square root of the average amplitude of the sEMG signal within a certain time window. Generally, when the muscle is subjected to a large load or fatigue, the RMS will decrease, and when the muscle is under a lighter load or is more relaxed, the RMS will increase.

[0070] The formula is:

[0071] Where N is the number of sEMG data points, x i is the i-th data in the signal data sequence.

[0072] Mean power frequency (MPF) refers to the average value of the power frequency distribution of the electromyographic signal generated during muscle contraction, which can be used to evaluate the strength and fatigue of muscle contraction. Even if there is some interference noise in the signal, this feature can better identify the useful information in the signal.

[0073] The formula is:

[0074] Where f is the frequency, and PSD(f) is the power spectrum of the current EMG data.

[0075] The median frequency (MF) refers to the median frequency of the electromyographic signal generated during muscle contraction, and is often used to assess the strength and fatigue of muscle contraction. Generally, when the muscle is under a greater load or fatigue, the MF will decrease, and when the muscle is under a lighter load or is more relaxed, the MF will increase. Its calculation formula is as follows:

[0076]

[0077] S3. Use the frost optimization algorithm to cyclically iterate the long short-term memory network parameters and construct a long short-term memory network model with optimal parameters.

[0078] like Figure 5 As shown in the figure, the decomposed intrinsic mode components are reconstructed into a data set and the LSTM model parameters are initialized. The model includes a memory unit and three gates (input gate, forget gate, and output gate).

[0079] Forget gate: determines which information in the memory unit needs to be forgotten.

[0080] The formula is: t =σ(W f ·[h t-1 ,x t ]+b f ).

[0081] Input gate: determines how the current input information affects the memory cell.

[0082] The formula is: t =σ(W f ·[h t-1 ,x t ]+b i ).

[0083] Candidate memory: Generate candidate memory for updating memory cells.

[0084] The formula is:

[0085] Update memory cell: Update the memory cell according to the output of the forget gate and the input gate.

[0086] The formula is:

[0087] Output gate: determines what information to output as the hidden state at the current moment.

[0088] The formula is: t =σ(W o ·[h t-1 ,x t ]+b o ).

[0089] Formula is o t =σ(W o ·[h t-1 ,x t ]+b o ).

[0090] By judging whether it is the optimal parameter, decide whether to perform LSTM prediction. If it is not the optimal parameter, optimize the parameter through RIME algorithm. Initialize the parameters to form random frost particles, iterate and optimize the frost particles, and detect whether the particles are in the optimal position. If they are in the optimal position, output the optimal parameter, otherwise continue iterative optimization.

[0091] The particle update calculation formula is:

[0092]

[0093] Where:

[0094] S4. Use the constructed long short-term memory network model to predict the various decomposition modes of the surface electromyography signal.

[0095] The optimized long short-term memory network model is constructed using the optimal parameters, and each modal component after reorganization and normalization is predicted at Δt time to obtain the modal component IMF after Δt time. 1 (T+Δt), IMF 2 (T+Δt)……IMF 3 (T+Δt).

[0096] For each modal component IMF 1 (T+Δt), IMF 2 (T+Δt)……IMF 3 (T+Δt) is superimposed according to the coefficient to obtain the predicted surface electromyographic signal.

[0097] S5. Construct a three-dimensional electromyographic adaptive control system based on the predicted signal as feedback.

[0098] like Figure 6 As shown, the surface electromyographic signal collected from the affected hand is processed to obtain characteristic values ​​such as integrated electromyographic value, root mean square value, average power frequency, and median frequency. The initial fatigue degree is calculated by the following formula:

[0099] x 1 =k 1 MPF+k 2 MF

[0100]

[0101] In the formula, x 1is the initial fatigue quantization value in the frequency domain, x 2 is the initial fatigue quantization value in the time domain. Coefficient k 1 To calculate the proportion of MPF in fatigue level, k 2 is the proportion of MF, RMS k The root mean square value in the kth time window. The initial muscle activation state is calculated by the following formula:

[0102]

[0103] Where y is the initial muscle activation degree, ω 1 is the proportion of integrated EMG values ​​below the average activation threshold, ω 2 is the proportion of integrated EMG values ​​above the average activation threshold, iEMG1 is the integrated EMG value below the average activation threshold, and iEMG2 is the integrated EMG value above the average activation threshold. The threshold is selected as 0.4 times the average integrated EMG value.

[0104] The surface electromyography collected from the affected hand is decomposed by CEEMDAN, and the modal data set is reconstructed. The LSTM model with optimal parameter optimization is used to predict the Δt time to obtain the modal component IMF after Δt time. 1 (T+Δt), IMF 2 (T+Δt)……IMF 3 (T+Δt).

[0105] The predicted surface electromyographic signal is feature extracted and a three-dimensional model based on the predicted surface electromyographic signal as feedback is constructed. The median frequency MF, mean power frequency MPF, integrated electromyographic value iEMG, and electromyographic root mean square value RMS are extracted respectively to obtain the following data:

[0106]

[0107] The three-dimensional model (X, Y, Z) is constructed by the following formula:

[0108]

[0109]

[0110] In the formula, X f is the fatigue degree change in the frequency domain, X Am is the change of fatigue degree in time domain, n is the number of modes, Y is the degree of muscle activation, Z is the muscle activation state, iEMGA is the integrated electromyographic value of the agonist muscle under the corresponding action, and iEMGB is the integrated electromyographic value of the antagonist muscle under the corresponding action. The three-dimensional model (X, Y, Z) of the predicted muscle is calculated by the above formula, and the system adaptively adjusts the parameter changes according to the feedback formed by the predicted muscle state.

[0111] S6. Generate fuzzy control rules based on adaptive fuzzy control system and electromyographic feedback.

[0112] like Figure 7 As shown, according to different predicted muscle states, the three-dimensional model (X, Y, Z) is used as input parameters. The change in fatigue level corresponds to different parameter adjustments of functional electrical stimulation and rehabilitation manipulator.

[0113] When X increases, that is, the patient's muscle fatigue increases, by reducing the working frequency of the rehabilitation robot and reducing the voltage, current intensity and stimulation frequency of functional electrical stimulation to slow down muscle fatigue and reduce risks.

[0114] When X decreases, that is, the patient's muscle fatigue level decreases, the rehabilitation intensity is increased by increasing the working frequency of the rehabilitation manipulator and increasing the voltage, current intensity and stimulation frequency of functional electrical stimulation, so as to increase the muscle activation and improve the rehabilitation effect.

[0115] When Y increases, that is, the patient's muscle activation level increases, and the patient's muscles are in good condition. By increasing the working frequency of the rehabilitation manipulator, the patient can complete the movement better, and reducing the voltage, current intensity and stimulation frequency of functional electrical stimulation to reduce muscle fatigue, so as to better adapt to the current muscle state.

[0116] When Y decreases, that is, the patient's muscle activation level is low, and the patient's muscle performance is poor. By reducing the working frequency of the rehabilitation robot, only providing movement assistance, and increasing the voltage, current intensity and stimulation frequency of functional electrical stimulation, the nerves can be better actively activated.

[0117] When Z = 1, the agonist and antagonist muscles are activated together, and the integrated electromyography of the two muscles produces almost the same fluctuation. This is often because the muscle tension of stroke patients is abnormally increased and spasticity occurs. At this time, the main focus should be on relieving spasticity and stopping the corresponding working state. When Z = 0, the agonist and antagonist muscles are not activated together, and the integrated electromyography of the two muscles does not fluctuate at the same time. At this time, rehabilitation training can continue.

[0118] S7. Control the combined rehabilitation system based on the control strategy.

[0119] like Figure 8As shown in the figure, the three-dimensional input (X, Y, Z) of the predicted state is fuzzified by the fuzzy controller, and the fuzzy control table constructed according to the three-dimensional electromyographic fuzzy rules is used for fuzzy decision making. The required output result is obtained after fuzzy reasoning, and it is defuzzified. According to the adaptive rules, it is judged whether the current muscle state matches the rehabilitation system parameters, and the system is adaptively adjusted. Finally, the parameters (F, U, I, H) are obtained, and the rehabilitation manipulator and the functional electrical stimulation system parameters are adjusted to match the subsequent muscle state. Among them, F represents the movement frequency of the rehabilitation manipulator, U represents the electrical stimulation voltage, I represents the electrical stimulation current, and H represents the electrical stimulation frequency.

[0120] Therefore, the present invention adopts the above-mentioned rehabilitation manipulator based on electromyographic feedback and electrical stimulation adaptive control method, which can ensure the safety of the rehabilitation process, reduce the possibility of muscle damage, help improve the rehabilitation training effect, and accelerate the recovery of neuroplasticity function and the rehabilitation process.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A rehabilitation manipulator based on electromyographic feedback and an adaptive control method for electrical stimulation, characterized in that: The following steps are involved: S1. Construct a myoelectric acquisition-rehabilitation manipulator-neuromuscular electrical stimulation combined rehabilitation system; S2, using complete set empirical mode decomposition and adaptive noise algorithm to preprocess the surface electromyography signal; S3, using the frost optimization algorithm to cyclically iterate the long short-term memory network parameters and construct a long short-term memory network model with optimal parameters; S4, using the constructed long short-term memory network model to predict each decomposition mode of the surface electromyography signal; S5, constructing a three-dimensional myoelectric adaptive control system based on the predicted signal as feedback; S6, Generate fuzzy control rules based on adaptive fuzzy control system and electromyographic feedback; S7. Control the combined rehabilitation system based on the control strategy.

2. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 1, characterized in that: In S1, the electromyographic signal is collected by electromyographic acquisition equipment, and the collected electromyographic signal is uploaded to the host computer for processing and analysis; the RS485 one-to-two serial port communication is used to establish communication between the host computer and the rehabilitation manipulator and functional electrical stimulation, and the rehabilitation manipulator is used to provide passive rehabilitation of the hand; active activation is provided through functional electrical stimulation.

3. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 1, characterized in that: In S2, DC component detection is used to remove the DC component, an adaptive 50Hz notch filter is used to remove power frequency interference, and fourth-order Butterworth high- and low-pass filters are used to remove high- and low-frequency noise in the signal. The processed surface electromyography signal is decomposed into modal components of different frequency bands using complete set empirical mode decomposition and adaptive noise model.

4. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 3, characterized in that: Using complete set empirical mode decomposition and adaptive noise model, the processed surface electromyography signal is decomposed from high to low to obtain the corresponding intrinsic mode components IMF1, IMF2...IMF n ; Decompose the one-dimensional surface electromyography signal data into multi-dimensional data in different frequency ranges so that the eigenvalues ​​fall within different frequency ranges. During the decomposition process, the first-order IMF component is calculated on an overall average basis to obtain the final first-order IMF component, and then the above operation is repeated on the residual part.

5. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 1, characterized in that: In S3, the intrinsic mode components of each surface electromyographic signal are reorganized and normalized to form a long short-term memory network model input data set; The initialized long short-term memory network model parameters constitute random frost particles, and the frost particles are iteratively optimized according to the particle update calculation formula to detect whether the particles are in the optimal position. If they are in the optimal position, the output is the optimal parameter, otherwise iterative optimization continues. The particle update calculation formula is:

6. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 1, characterized in that: In S4, the optimized long short-term memory network model is constructed using the optimal parameters, and each modal component after reorganization and normalization is predicted at time Δt to obtain the modal components IMF1(T+Δt), IMF2(T+Δt)…IMF3(T+Δt) after time Δt; Each modal component IMF1(T+Δt), IMF2(T+Δt)…IMF3(T+Δt) is superimposed according to the coefficient to obtain the predicted surface electromyography signal.

7. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 1, characterized in that: In S5, the predicted surface electromyographic signal is feature extracted and a three-dimensional model based on the predicted surface electromyographic signal as feedback is constructed; the median frequency MF, the mean power frequency MPF, the integrated electromyographic value iEMG, and the electromyographic root mean square value RMS are extracted respectively. The three-dimensional model (X, Y, Z) is constructed by the following formula: Where, X f is the fatigue degree change in the frequency domain, X Am is the change of fatigue degree in time domain, n is the number of modes, Y is the degree of muscle activation, Z is the muscle activation state, iEMGA is the integrated EMG value of the agonist muscle under the corresponding action, and iEMGB is the integrated EMG value of the antagonist muscle under the corresponding action; the system adaptively adjusts parameter changes according to the feedback formed by the predicted muscle state.

8. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 1, characterized in that: In S6, the predicted three-dimensional model is used as the input of the fuzzy rule: when the predicted muscle fatigue degree X increases, the movement frequency of the rehabilitation manipulator decreases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all decrease; when the predicted muscle fatigue degree X decreases, the movement frequency of the rehabilitation manipulator increases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all increase; when the predicted muscle activation degree Y increases, the movement frequency of the rehabilitation manipulator increases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all decrease; when the predicted muscle activation degree Y decreases, the movement frequency of the rehabilitation manipulator decreases, and the stimulation voltage, current, and frequency of the functional electrical stimulation all increase; when the predicted activation state Z is normal, the joint rehabilitation system continues to work; when the predicted activation state Z is abnormal, the joint rehabilitation system stops working.

9. The method for adaptive control of a rehabilitation manipulator and electrical stimulation based on myoelectric feedback according to claim 1, characterized in that: In S7, the three-dimensional input (X, Y, Z) of the predicted state is fuzzified by the fuzzy controller, and fuzzy decision is made according to the fuzzy control table constructed by the three-dimensional electromyography fuzzy rules. The required output result is obtained after fuzzy reasoning, and it is defuzzified. According to the adaptive rules, it is judged whether the current muscle state matches the rehabilitation system parameters, and the system is adaptively adjusted to finally obtain the parameters (F, U, I, H), where F represents the movement frequency of the rehabilitation manipulator, U represents the electrical stimulation voltage, I represents the electrical stimulation current, and H represents the electrical stimulation frequency; the parameters of the combined rehabilitation system are adjusted to match the subsequent muscle state.

Images