Lower limb rehabilitation robot control method and system based on electromyographic signals

Abstract

The invention discloses a lower limb rehabilitation robot control method and system based on electromyographic signals, and belongs to the technical field of medical rehabilitation instruments.The method comprises the steps that the electromyographic signals of lower limb muscles of all parts of a user when the user uses a lower limb rehabilitation robot are collected in real time, and the electromyographic signals are preprocessed; the electromyographic signal features of the electromyographic signals are input into the trained DE fused GA-SVM algorithm classification and recognition model, the gait cycle stage is determined, main force exerting muscles in the gait cycle stage are selected, and the muscle force level, needing to be compensated, of the main force exerting muscles is output. According to the method, the problem that the lower limb muscle activation levels of different gait phases are different is considered, the current gait cycle stage of a user is judged firstly, muscle force compensation is conducted on the muscles mainly exerting force at different stages, a patient can conduct active rehabilitation training, the muscle force training efficiency of the patient can be improved, and the rehabilitation training efficiency of the patient is improved. And a better rehabilitation training effect is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical rehabilitation equipment, and in particular to a lower limb rehabilitation robot control method and system based on electromyographic signals. Background Art

[0002] Lower limb rehabilitation robots can help patients with scientific and effective rehabilitation treatment, improve their limb motor ability, delay muscle atrophy and joint contracture, and effectively promote the functional reorganization of the nervous system. Lower limb rehabilitation robots are automated medical devices that enable patients to restore normal walking function through repeated exercise training of the lower limbs. They include a weight-reducing suspension system, a mobile treadmill, mechanical legs, and a control panel. The suspension system is used to provide patients with appropriate support, reduce some of their body weight, remove the weight that their lower limbs should bear and redistribute it, thereby reducing the burden on their legs. The mechanical legs are controlled through the control panel so that the mechanical legs drive the patient to walk on the mobile treadmill with a normal gait to prevent the formation of abnormal gait.

[0003] At present, the active control algorithms of lower limb rehabilitation robots have become increasingly perfect, and machine control systems have emerged that can promote the active rehabilitation training effects of patients to a certain extent. Enabling lower limb rehabilitation robots to provide targeted training for different patients is the key to improving the effectiveness of rehabilitation training. For example, the Chinese invention patent application with publication number CN117398264A and publication date 2024.01.16, entitled A lower limb rehabilitation system and control method capable of automatically switching active control modes, discloses a gait recognition table that specifies the correspondence between plantar pressure data, motion posture data, and interaction force data between the human body and the exoskeleton and gait stages, and judges the gait stage according to the collected plantar pressure data, motion posture data, and interaction force data between the human body and the exoskeleton, and through a stage muscle control table that specifies the correspondence between the gait stage and whether each part of the muscle is contracted, performs abnormal judgment on the electromyographic signals obtained in the corresponding gait stage, and performs different controls based on different results of the abnormal judgment. Therefore, this Chinese invention patent application enables the lower limb rehabilitation exoskeleton robot system to have the functions of real-time identification of electromyographic signal abnormalities and real-time adjustment of responses, avoiding the problem of abnormal exoskeleton movement execution caused by insufficient muscle function recovery and unstable electromyographic signal acquisition, and thus, for the situation where the muscle function of different parts of different patients is incompletely recovered, there is no need for personalized processing, and it can adapt to patients with different degrees of muscle recovery.

[0004] However, the Chinese invention patent application needs to obtain plantar pressure data, motion posture data, and interaction force data between the human body and the exoskeleton when judging the gait stage. The acquisition of this data requires the addition of corresponding sensors on the lower limb rehabilitation robot and connecting it to the corresponding parts of the user, resulting in an increase in the detection equipment on the rehabilitation robot and a corresponding increase in the acquisition sensors on the user, which increases the weight during the training process and inevitably affects the training effect of the user. In addition, the prior art judges the abnormality of the acquired electromyographic signal according to the stage muscle control table, and controls the overall training of the patient's lower limbs under different judgment results, but does not control the lower limb rehabilitation robot accordingly based on the problem that the lower limb muscle activation levels of the patient in different gait phases during rehabilitation training are different, resulting in the patient's muscle strength not being well targeted during rehabilitation training. Therefore, without adding additional acquisition equipment, the problem of controlling the lower limb rehabilitation robot so that different patients can receive targeted training has become an urgent problem to be solved in the production process of enterprises. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a lower limb rehabilitation robot control method and system based on electromyographic signals, so as to solve the problem that the existing process of data collection not only requires the addition of an electromyographic signal acquisition module but also requires the addition of multiple categories of additional acquisition devices such as plantar pressure data, motion posture data, and interaction force data between the human body and the exoskeleton. The increase in the types of data collected will lead to the complexity of data processing, and there is no targeted training for the lower limb muscles in different gait phases, resulting in the problem that the patient's muscle strength is not well targeted during rehabilitation training.

[0006] To achieve the above object, the present invention provides a lower limb rehabilitation robot control method based on electromyographic signals. Specifically, the method comprises the following steps:

[0007] 1) Obtaining the electromyographic signals of the lower limb muscles of various parts of the user when the user uses the lower limb rehabilitation robot for training, output by the data acquisition module;

[0008] 2) The acquired electromyographic signal is transmitted to a preprocessing module to preprocess the acquired electromyographic signal to obtain an electromyographic signal after removing power frequency interference and denoising;

[0009] 3) Extract the electromyographic signal features of the preprocessed electromyographic signal that can distinguish the gait cycle stage of the muscle, input the electromyographic signal features into the trained DE fused GA-SVM algorithm classification and recognition model, judge the gait cycle stage, and according to the judgment result, select the sampling device connected to each lower limb muscle part corresponding to the movement of the gait cycle stage, obtain the electromyographic signals of each lower limb muscle part of the movement, determine the muscle strength level of each lower limb muscle part at this time according to the electromyographic signals of each lower limb muscle part, and control the driver to perform muscle strength compensation on the lower limb muscle part that needs compensation according to the muscle strength level.

[0010] Its beneficial effects are as follows: the method of the present invention takes into account the problem that the activation levels of lower limb muscles in different gait phases of patients during rehabilitation training are different. By first judging the current gait cycle stage of the user, and performing muscle strength compensation for each lower limb muscle part corresponding to the movement of the gait cycle stage (i.e., performing muscle strength compensation for the main muscles at different stages), the patient can be enabled to perform active rehabilitation training, and the training efficiency of the patient's muscle strength can be improved, so as to achieve better rehabilitation training effect. And the method of the present invention can judge the gait cycle stage of the user based on the collection and processing of electromyographic signals and the use of deep learning methods, and the data type obtained by the acquisition device of electromyographic signal acquisition is single, so the complexity of data processing is reduced, and then the accuracy of data processing can be correspondingly improved, and the electromyographic signal acquisition in the present invention only adds electromyographic signal acquisition equipment (for example, attaching electrode sheets on the muscle surface to connect input electrodes) to realize electromyographic signal acquisition, avoiding the process of data acquisition of the user's training by the lower limb rehabilitation robot with too many additional types of acquisition equipment, resulting in an increase in the weight load during the user's training process, affecting the user's training effect.

[0011] Based on the above, in step 3), the training steps of the GA-SVM algorithm classification and recognition model fused with DE include:

[0012] 3.1) Collect multiple groups of lower limb electromyographic signals of lower limb muscles of various parts of the human body under different motion states as the initial data set, and pre-process the initial data set to remove power frequency interference and denoise to obtain the pre-processed lower limb electromyographic signals;

[0013] 3.2) Dividing the preprocessed lower limb electromyographic signals into gait cycle stages, extracting electromyographic signal features of the preprocessed lower limb electromyographic signals that can distinguish the gait cycle stages of the muscles, and constructing an electromyographic signal feature training set;

[0014] 3.3) Use the electromyography signal feature training set to train the DE fused GA-SVM algorithm classification and recognition model.

[0015] The method of the present invention uses a feature type of electromyographic signal characteristics that can distinguish the gait cycle stages of muscles for training when training the model. Therefore, in actual use, when the user performs rehabilitation training through a lower limb rehabilitation robot, the electromyographic signal characteristics that can distinguish the gait cycle stages of muscles are also used as input, so that the input algorithm classification model in the actual application process and the input algorithm classification model in the training model process are the same type of information, so the accuracy of the results obtained in the actual application process is also guaranteed.

[0016] Based on the above, in step 3.2), the pre-processed lower limb electromyographic signals are divided into gait cycle stages by defining energy thresholds and using a sliding window method.

[0017] The present invention defines an energy threshold and applies it to a sliding window method to effectively identify and extract the parts of lower limb muscle movements in different gait phases (i.e. different gait cycle stages), thereby achieving accurate recognition of the action.

[0018] Based on the above, in step 3), the method for determining the feature type of the electromyographic signal characteristics that can distinguish the gait cycle stage of the muscles is as follows: obtain the electromyographic signal feature information of multiple feature types of the electromyographic signal after the gait cycle stage division in step 3.2), analyze the change differences of the electromyographic signal feature information of multiple feature types between different gait cycle stages, select a specific number of feature types whose difference reaches a preset standard according to the size of the change difference, and determine this specific number of feature types as the feature type of the electromyographic signal characteristics that can distinguish the gait cycle stage of the muscles.

[0019] The method of the present invention can effectively select those features that have a strong influence on the target variable from the original feature set through feature selection, while excluding those features that have a small influence or no influence. This process not only helps to reduce the complexity of the model, but also provides a basis for subsequent feature fusion, further enhancing the model's ability to understand and interpret data.

[0020] Based on the above, in step 3), after obtaining the electromyographic signal features that can distinguish the gait cycle stages of the muscles, the GASF algorithm and the LDA algorithm are used in turn to enhance the features of the electromyographic signal, and then the enhanced electromyographic signal features are input into the DE fused GA-SVM algorithm classification and recognition model.

[0021] The present invention adopts a method combining GASF and LDA. First, GASF is used to convert the time series data of electromyographic signals into images, and then the LDA method is used to reduce the dimension to find the features of the linear combination of electromyographic signals, and the electromyographic signal features with discriminability are extracted, which provides a basis for the effective classification of the trainer.

[0022] Based on the above, in step 3), the method for determining the muscle strength level of each lower limb muscle part at this time according to the electromyographic signals of each lower limb muscle part is: according to the electromyographic signals of each lower limb muscle part, the proportion of energy consumed by each lower limb muscle part is obtained, and according to the proportion of energy consumed by each lower limb muscle part and the proportion of energy consumed by each lower limb muscle part under normal conditions in the set gait cycle stage, the muscle strength level of the lower limb muscle part at this time is determined.

[0023] Since the proportion of energy consumed by each muscle in each phase of a complete gait cycle is basically the same, the proportion of energy consumed by different muscles in each phase of the complete gait cycle can be compared with the normal energy proportion. Muscles that consume less energy than normal are muscles with lower activation levels. Therefore, based on this, the muscle strength level of each muscle part can be determined, and then targeted training can be performed on the muscle part according to the muscle strength level.

[0024] Based on the above, in step 3), the lower limb muscle parts that need to be compensated are: the lower limb muscle parts whose proportion of energy consumed is lower than the proportion of energy consumed by the lower limb muscle parts in the normal state of the set gait cycle stage and the corresponding lower limb muscle parts in the movement.

[0025] Because the muscle strength level in the present invention is determined based on the energy ratio, the muscle parts of the lower limbs whose energy consumption ratio is lower than the normal value can be judged according to the muscle strength level, and the lower limb muscles in this part are the lower limb muscles that need muscle strength compensation. Therefore, muscle strength compensation for this muscle part can maximize the user's muscle strength.

[0026] Based on the above, the gait cycle stages include the early support period from heel strike to contralateral lift-off, the middle support period from contralateral lift-off to midpoint of support, the late support period from midpoint of support to contralateral landing, the early swing period from contralateral landing to toe-off, and the late swing period from toe-off to heel strike.

[0027] The method of the present invention takes into account the problem that the muscle activation states of different gait phases in a complete gait cycle are different. Therefore, the complete gait cycle is divided into the above-mentioned five stage categories according to the differences in muscle activation states, and then the gait cycle stage can be accurately determined based on the muscle electromyographic signals of each part.

[0028] Based on the above, in step 1), the electromyographic signal is a signal output after the acquisition module collects the original electromyographic signals of the lower limb muscles of various parts of the user when using the lower limb rehabilitation robot for training, and then amplifies and filters the original electromyographic signals in the operational amplifier and the filter.

[0029] The present invention amplifies and filters the original signal, and can amplify the collected weak voltage, current or charge signal through the amplifier to increase its amplitude and power for subsequent processing and analysis, and extract useful signals through the filter to suppress unnecessary interference, making the system more stable.

[0030] To achieve the above-mentioned objectives, the present invention also provides a lower limb rehabilitation robot control system based on electromyographic signals, including a data acquisition module for collecting electromyographic signals of various parts of the user's lower limb muscles when using the lower limb rehabilitation robot for training, a preprocessing module for preprocessing the acquired electromyographic signals, and an electromyographic signal classification control module. The electromyographic signal control module is used to execute instructions to implement the steps of the lower limb rehabilitation robot control method and achieve the same beneficial effects as the lower limb rehabilitation robot control method.

[0031] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the following specifically cites a preferred embodiment and describes it in detail with the accompanying drawings as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 Block diagram of the control system of the lower limb rehabilitation robot based on electromyographic signals of the present invention.

[0033] Figure 2 Schematic diagram of the human gait cycle of the present invention.

[0034] Figure 3 Flow chart of surface electromyography signal feature extraction of the present invention.

[0035] Figure 4 Variable power assist control flow chart of the present invention. DETAILED DESCRIPTION

[0036] The technical scheme of the present invention will be clearly and completely described below in conjunction with specific embodiments, but those skilled in the art should understand that the embodiments described below are only used to illustrate the present invention and should not be regarded as limiting the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Embodiment of lower limb rehabilitation robot control system based on electromyographic signal

[0038] like Figure 1As shown, the lower limb rehabilitation robot control system of this embodiment includes a data acquisition module for collecting electromyographic signals of lower limb muscles of various parts of the user when using the lower limb rehabilitation robot for training, a preprocessing module for preprocessing the acquired electromyographic signals, and an electromyographic signal classification control module. The electromyographic signal control module obtains the gait cycle stage according to the preprocessed electromyographic signals, and performs muscle strength compensation for the main force-generating muscles at different stages, so as to enable the patient to perform active rehabilitation training and improve the patient's muscle strength training efficiency, thereby achieving better rehabilitation training effects.

[0039] The system of this embodiment includes an early training process and a later application process. Specifically, the early training process includes:

[0040] 1) Signal acquisition.

[0041] Multiple groups of human lower limb electromyographic signal collection were performed, and surface electromyographic signals of different muscles of the subjects were collected in standing state, normal walking state and weight-reduced walking state, that is, the collected signals were signals of the complete movement process of each part.

[0042] First, the data acquisition module is used for data acquisition. In this embodiment, based on the theory of rehabilitation medicine, the five muscles most closely related to the lower limb movement of the human body are selected for signal acquisition, and the selected muscles are the rectus femoris, vastus lateralis, biceps femoris, soleus and gastrocnemius. The subjects started lower limb exercise in a weight loss environment and a normal environment respectively and collected lower limb electromyographic signals in a complete gait cycle in real time. The electrode sheet is attached to the surface of the muscle to connect the input electrode to transmit the collected signal to the operational amplifier and filter for amplification and filtering. After processing, the output signal is output to the computer via WiFi to generate a signal curve. The working voltage of the acquisition equipment is 9V, the sampling frequency is 1000Hz, and the voltage signal amplification factor is 1000 times.

[0043] 2) Signal preprocessing.

[0044] The initial signal was preprocessed, and the 50HZ power frequency interference was filtered out by a notch filter, and the noise and artifacts in the signal were removed by soft threshold wavelet denoising. The surface electromyography signal was automatically segmented by the sliding window energy threshold method, and the fixed thresholds of the start and end points of the gesture interval were calculated. By defining the energy threshold and applying it to the continuously sliding window, the complete gait cycle of the collected signal was divided into 5 categories, and each phase category accounted for 20% of the complete gait cycle. The lower limb muscle movement part was identified and extracted, thereby achieving accurate recognition of the action.

[0045] Specifically, in this embodiment, the preprocessing of the electromyographic signals of the lower limbs is performed based on the electromyographic signals of the five muscles, namely, the rectus femoris, vastus lateralis, biceps femoris, soleus and gastrocnemius, collected in the previous steps at different gait phases. First, the frequency spectrum characteristics of the electromyographic signals are analyzed by the frequency domain analysis method to determine the type of muscle contraction. Next, the time characteristics of the electromyographic signals are analyzed by the time domain analysis method to achieve the effect of judging the intensity and duration of muscle contraction. Since the 50HZ power frequency has a great interference on the FFT transformation of the original electromyographic signal, an IIR digital notch filter is now used to obtain the signal after the notch. After eliminating the power frequency interference, the main frequency of the electromyographic signal is concentrated in 0-500HZ, especially in 10-300HZ. In addition, the soft threshold wavelet is also used to denoise the electromyographic signal. It can effectively remove the noise in the electromyographic signal and improve the signal-to-noise ratio of the electromyographic signal. At the same time, the sliding window energy threshold method is used to segment the electromyographic signal, and a complete gait cycle is divided into the following: Figure 2 The lower limb electromyographic signals of the five gait phases shown in the figure are designed to remove invalid or redundant parts of the signal and retain only valid signals related to the lower limb muscle movements. In this way, the interference of invalid data on the recognition system can be reduced, and the problems of low recognition rate and excessive data processing caused by improper data processing can be reduced. The core of this method is to effectively identify and extract the parts of the lower limb muscle movement by defining the energy threshold and applying it to the continuously sliding window, thereby achieving accurate recognition of the action and muscle strength classification.

[0046] This embodiment takes into account the problem that the muscle activation states of different gait phases within a complete gait cycle are different. By defining an energy threshold and applying it to a sliding window, the parts of the lower limb muscle movements in five different gait phases are effectively identified and extracted, thereby achieving accurate recognition of the movements.

[0047] 3) Construct a DE fused GA-SVM algorithm classification and recognition model and train the model.

[0048] Extract the features of the electromyographic signal after automatic segmentation. By analyzing the time domain, frequency domain and time-frequency domain of the electromyographic signal, select the root mean square (RMS), mean absolute value (MAV), variance (VAR), integrated voltage (IEMG), power spectral density (PSD), median frequency (MF), mean power frequency (MPF) and wavelet coefficient as the extracted features. These features are actually classified and compared for analysis, and the optimal features are screened and the suitable features are fused to construct the optimal feature set. Combined with GASF and LDA to enhance the features of the electromyographic signal, GASF (Gramian Angular Field) can be used to convert the time series data into an image, and LDA (Linear Discriminant Analysis) can be used to find a linear combination of features. When processing the image converted by GASF, LDA can be used to extract the features that can best distinguish different categories, and these features are used to train the classifier of the constructed DE fusion GA-SVM algorithm classification and recognition model for effective classification.

[0049] Specifically, Figure 3 As shown, this embodiment extracts features from the processed electromyographic signal, and through analysis of the time domain, frequency domain and time-frequency domain of the electromyographic signal, selects the root mean square (RMS), mean absolute value (MAV), variance (VAR), integrated voltage (IEMG), power spectral density (PSD), median frequency (MF), mean power frequency (MPF) and wavelet coefficients as extracted features, and actually classifies and compares these features, performs optimal feature screening and fuses suitable features to construct an optimal feature set. Through the selection of these features, those features that have a strong influence on the target variable can be effectively selected from the original feature set, while those features with little or no influence can be excluded. This process not only helps to reduce the complexity of the model, but also provides a basis for subsequent feature fusion, further enhancing the model's ability to understand and interpret data. As Figure 4 As shown in the figure, a feature enhancement method based on the GASF-LDA algorithm is used to reduce the dimension and extract features of the root mean square (RMS), mean absolute value (MAV), variance (VAR), integrated voltage (IEMG), power spectral density (PSD), median frequency (MF), mean power frequency (MPF) and wavelet coefficients, thereby improving the performance of classification or regression tasks. Using the PyTorch deep learning framework, a GA-SVM algorithm classification and recognition model based on DE fusion is constructed. The model uses the EMG signal features after dimensionality reduction as input into the first classifier, and the output is the categories of five gait phases.

[0050] This embodiment adopts a method combining GASF and LDA. First, GASF is used to convert the time series data of the electromyographic signal into an image, and then the LDA method is used to reduce the dimension to find the features of the linear combination of the electromyographic signal, and the electromyographic signal features with discriminability are extracted, providing a basis for the effective classification of the trainer.

[0051] Based on the previous training process of the above system, in actual application, the system of this embodiment implements muscle strength compensation for the main muscles at different stages through the following steps:

[0052] 1) Real-time signal acquisition: The data acquisition module is used to obtain the electromyographic signals of the lower limb muscles of various parts of the user when the user is using the lower limb rehabilitation robot for training. In this embodiment, the electromyographic signals are signals output after the acquisition module acquires the original electromyographic signals of the lower limb muscles of various parts of the user when the user is using the lower limb rehabilitation robot for training, and the original electromyographic signals are amplified and filtered in the operational amplifier and the filter.

[0053] 2) Signal preprocessing: The acquired electromyographic signal is transmitted to the preprocessing module to preprocess the acquired electromyographic signal to obtain the electromyographic signal after removing the power frequency interference and denoising, so as to achieve the effect of filtering out the 50HZ power frequency interference and removing the noise in the electromyographic signal.

[0054] 3) Signal classification control: extract the electromyographic signal features of the preprocessed electromyographic signal that can distinguish the gait cycle stage of the muscle, input the electromyographic signal features into the trained DE fused GA-SVM algorithm classification and recognition model, obtain the gait cycle stage, select the sampling device connected to each lower limb muscle part corresponding to the movement of the gait cycle stage, obtain the electromyographic signals of each lower limb muscle part of the movement, determine the muscle strength level of each lower limb muscle part at this time according to the electromyographic signals of each lower limb muscle part, and control the driver to perform muscle strength compensation on the lower limb muscle part that needs compensation according to the muscle strength level.

[0055] This embodiment adopts deep learning to construct a GA-SVM algorithm classification and recognition model based on DE fusion. During the model training process, the gait phase is first judged to find the current main force-generating muscles. The muscle signal input weights are used as input signals to enter the classifier. The classifier then selects the optimal channel for the assistance level that needs to be compensated, thereby realizing variable assistance muscle strength compensation and achieving the effect of greatly improving the patient's muscle strength.

[0056] Specifically, in this embodiment, the signal is collected and analyzed through a sliding window, and the extracted features are provided to the GA-SVM algorithm classification and recognition model of DE fusion for gait phase judgment, and the optimal channel is selected according to the phase to judge the power level that needs to be compensated. The classification result of the electromyographic signal output by the classifier is input into the motion controller, and the motion controller fuses the torque sensor and the angle sensor to determine the motion trajectory. The PID control of the multi-fusion signal is used to adjust the deviation between the actual trajectory and the expected trajectory in real time, that is, in this embodiment, the gait cycle stage is obtained according to the GA-SVM algorithm classification and recognition model of DE fusion after training, and the electromyographic signal of the main channel is selected according to the gait cycle stage, and the features of the electromyographic signals of these main channels are extracted, and then the features are input into the muscle strength level classifier to obtain the output muscle strength level signal, and then the muscle strength level signal is input into the robot controller for corresponding power compensation. After the compensation power required in different phases is determined by the model, the driver response is controlled to achieve variable power compensation under different gait phases, so that the muscles are more fully trained.

[0057] In this embodiment, since the energy consumption ratio of each muscle in each phase of a complete gait cycle is basically the same, the energy consumption ratio of different muscles in each phase of the complete gait cycle of the subject can be compared with the normal energy ratio, and the muscles with energy consumption ratio lower than the normal value are muscles with low activation levels. In addition, the energy consumption ratio of each muscle in different gait phases can be used as input to construct an optimal classification model, and different muscle strength levels can be identified according to the energy consumption ratio, and different muscle strength level results can be used as input signals for the controller of the lower limb rehabilitation robot.

[0058] This embodiment provides a lower limb robot control system that has different characteristics of muscle activation levels under different gait phases in the same gait cycle and phase-adjusts the power according to the different gait of the subject. The main method of deep learning is used to realize the judgment of the compensation power level, so as to adapt to patients with different gait phases for effective variable power rehabilitation training. The surface electromyographic signal is automatically segmented by the sliding window energy threshold method, and the five gait phases of lower limb muscle movement are effectively segmented. After collecting sufficient electromyographic signal data, the root mean square (RMS), mean absolute value (MAV), variance (VAR), integrated voltage (IEMG), power spectral density (PSD), median frequency (MF), mean power frequency (MPF) and wavelet coefficients are selected as extracted features, and the electromyographic signal features are enhanced by the method of combining GASF and LDA, thereby improving the level classification performance of power compensation. Finally, the GA-SVM algorithm classification and recognition model based on DE fusion is used for classification and recognition, gait phase judgment is performed, and the optimal compensation power level is selected according to the current phase to achieve the control effect of variable power compensation rehabilitation. Compared with the existing technology, this method takes into account the different levels of muscle activation in the gait cycle, which enables different muscles in the gait cycle to be more fully trained, and provides an innovative solution for the control of lower limb rehabilitation devices.

[0059] Embodiment of lower limb rehabilitation robot control method based on electromyographic signal

[0060] The lower limb rehabilitation robot control method of this embodiment is implemented by a lower limb rehabilitation robot control system based on electromyographic signals, and the method steps for implementing the lower limb rehabilitation robot control system have been introduced in detail in the embodiment of the lower limb rehabilitation robot control system based on electromyographic signals, and will not be repeated here.

[0061] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A lower limb rehabilitation robot control method based on electromyographic signals, characterized in that: The steps include: 1) Obtaining the electromyographic signals of the lower limb muscles of various parts of the user when the user uses the lower limb rehabilitation robot for training, output by the data acquisition module; 2) The acquired electromyographic signal is transmitted to a preprocessing module to preprocess the acquired electromyographic signal to obtain an electromyographic signal after removing power frequency interference and denoising; 3) Extract the electromyographic signal features of the preprocessed electromyographic signal that can distinguish the gait cycle stage of the muscle, input the electromyographic signal features into the trained DE fused GA-SVM algorithm classification and recognition model, judge the gait cycle stage, and according to the judgment result, select the sampling device connected to each lower limb muscle part corresponding to the movement of the gait cycle stage, obtain the electromyographic signals of each lower limb muscle part of the movement, determine the muscle strength level of each lower limb muscle part at this time according to the electromyographic signals of each lower limb muscle part, and control the driver to perform muscle strength compensation on the lower limb muscle part that needs compensation according to the muscle strength level.

2. The lower limb rehabilitation robot control method according to claim 1, characterized in that: In step 3), the training steps of the GA-SVM algorithm classification and recognition model fused with DE include: 3.1) Collect multiple groups of lower limb electromyographic signals of lower limb muscles of various parts of the human body under different motion states as the initial data set, and pre-process the initial data set to remove power frequency interference and denoise to obtain the pre-processed lower limb electromyographic signals; 3.2) Dividing the preprocessed lower limb electromyographic signals into gait cycle stages, extracting electromyographic signal features of the preprocessed lower limb electromyographic signals that can distinguish the gait cycle stages of the muscles, and constructing an electromyographic signal feature training set; 3.3) Use the electromyography signal feature training set to train the DE fused GA-SVM algorithm classification and recognition model.

3. The lower limb rehabilitation robot control method according to claim 2, characterized in that: In step 3.2), the preprocessed lower limb electromyographic signals are divided into gait cycle stages by defining energy thresholds and using a sliding window method.

4. The lower limb rehabilitation robot control method according to claim 2, characterized in that: In step 3), the method for determining the feature type of the electromyographic signal characteristics that can distinguish the gait cycle stage of the muscles is as follows: obtain the electromyographic signal feature information of multiple feature types of the electromyographic signal after the gait cycle stage division in step 3.2), analyze the change differences of the electromyographic signal feature information of multiple feature types between different gait cycle stages, select a specific number of feature types whose difference reaches a preset standard according to the size of the change difference, and determine this specific number of feature types as the feature type of the electromyographic signal characteristics that can distinguish the gait cycle stage of the muscles.

5. The lower limb rehabilitation robot control method according to claim 2, characterized in that: In step 3), after obtaining the electromyographic signal features that can distinguish the gait cycle stages of the muscles, the GASF algorithm and the LDA algorithm are used in turn to enhance the features of the electromyographic signal features, and then the enhanced electromyographic signal features are input into the DE fused GA-SVM algorithm classification and recognition model.

6. The lower limb rehabilitation robot control method according to claim 1, characterized in that: In step 3), the method for determining the muscle strength level of each lower limb muscle part at this time according to the electromyographic signals of each lower limb muscle part is as follows: according to the electromyographic signals of each lower limb muscle part, the proportion of energy consumed by each lower limb muscle part is obtained, and according to the proportion of energy consumed by each lower limb muscle part and the proportion of energy consumed by each lower limb muscle part under normal conditions in the set gait cycle stage, the muscle strength level of the lower limb muscle part at this time is determined.

7. The lower limb rehabilitation robot control method according to claim 6, characterized in that: In step 3), the lower limb muscle parts that need to be compensated are: the lower limb muscle parts corresponding to the movement whose proportion of energy consumed is lower than the proportion of energy consumed by the lower limb muscle parts in the normal state of the set gait cycle stage.

8. The lower limb rehabilitation robot control method according to claim 1, characterized in that: The gait cycle phases include the early support period from heel strike to contralateral lift-off, the middle support period from contralateral lift-off to midpoint of support, the late support period from midpoint of support to contralateral touchdown, the early swing period from contralateral touchdown to toe-off, and the late swing period from toe-off to heel strike.

9. The lower limb rehabilitation robot control method according to claim 1, characterized in that: In step 1), the electromyographic signal is a signal output after the acquisition module acquires the original electromyographic signals of the lower limb muscles of various parts of the user when using the lower limb rehabilitation robot for training, and the original electromyographic signals are amplified and filtered in an operational amplifier and a filter.

10. A lower limb rehabilitation robot control system based on electromyographic signals, characterized in that: It includes a data acquisition module for collecting electromyographic signals of various parts of the user's lower limb muscles when using a lower limb rehabilitation robot for training, a preprocessing module for preprocessing the acquired electromyographic signals, and an electromyographic signal classification control module. The electromyographic signal control module is used to execute instructions to implement the steps of the lower limb rehabilitation robot control method described in any one of claims 1-9.

Images