Research On Gesture Recognition Of Disabled Hand Rehabilitation Robot Based On Variational Mode Decomposition And Support Vector Machine

The aging population and the increase of the disabled have led to an increasing demand for rehabilitation robots.However,the existing hand rehabilitation robots are mainly rigid in structure and bulky,and it is difficult to meet the daily rehabilitation training needs of patients.It is particularly important to develop a portable hand soft rehabilitation robot with multiple training methods.This thesis takes the self-developed hand software rehabilitation robot as the research object,and carries out the research on surface electromyography signal(s EMG)noise reduction,feature extraction and gesture recognition methods.The main research contents are as follows:First of all,according to the technical requirements of the soft hand rehabilitation robot,the overall scheme design is carried out,and the design schemes of the software rehabilitation control system,s EMG acquisition system,and software system of the hand soft rehabilitation robot are completed,and the production process of the soft glove is introduced.Then,the noise reduction method of s EMG signal is studied,and a s EMG signal noise reduction method based on adaptive variational mode decomposition algorithm(AVMD)is proposed.Aiming at the problem that the variational mode decomposition algorithm needs preset parameters,an artificial bee colony algorithm with improved objective function is used to adaptively select the variational modal decomposition parameters;aiming at the problem that the high-frequency components obtained by the decomposition of s EMG signals by AVMD algorithm still contain noise,an improved wavelet threshold noise reduction algorithm is used to reduce noise for high-frequency components,and the noise reduction components and other components of the wavelet threshold noise reduction algorithm are superimposed to obtain the s EMG signal after noise reduction,and the effectiveness of the noise reduction algorithm proposed in this thesis is verified by experiments.Secondly,the research on s EMG signal feature extraction and gesture recognition is carried out.A feature extraction method based on reconstruction independent component analysis(RICA)is proposed to solve the problem of inter-channel interference of multi-channel s EMG signals.RICA-VAR-Fuzzy En features are constructed by extracting variance(VAR)and fuzzy entropy(Fuzzy En)obtained by reconstruction independent variable analysis transformations of s EMG signals.Aiming at the multi-classification problem of gesture recognition,a multi-classifier based on twin support vector machine(TWSVM)is designed for classification recognition.The feasibility and effectiveness of the feature extraction method based on reconstruction independent component analysis and the designed multi-classification method based on twin support vector machine proposed in this thesis are verified by experiments.Finally,the hand software rehabilitation robot system is developed and related experiments and analysis are carried out.The development work mainly includes the control system of the software rehabilitation robot,the s EMG signal acquisition system,and the software system of the rehabilitation robot.After the debugging of the hand soft rehabilitation robot system,the feasibility of the noise reduction method based on adaptive variational modal decomposition,the feature extraction method based on reconstruction independent component analysis and the multi-classification method based on twin support vector machine proposed in this thesis are verified in the actual acquired s EMG signals.The algorithm is applied to the hand soft rehabilitation robot,and the joint debugging is carried out,and the s EMG signal monitoring experiment,passive training experiment and master-slave training experiment are carried out,and the experimental results show the effectiveness of the hand soft rehabilitation robot based on the algorithm in this thesis.