Research On Movement Recognition Based On Synergistic Myoelectrical Activities

The upper limb is an important motor function organ of the human body and plays an important role in daily life.Physical disability and upper limb motor dysfunction due to various reasons can have a great impact on the daily life of the patient and cause a huge psychological and physiological burden.The use of surface EMG signals to identify motor intent and control rehabilitation systems has always been a focus of attention in neuroengineering and rehabilitation engineering.The recognition of motion patterns based on the characteristics of myoelectric signals is the key link of myoelectric control.At present,the common methods of feature value extraction include time domain,frequency domain and time-frequency domain.Muscle synergistic contraction is a basic strategy of the nervous system for motion control.Supported by the intelligent interactive control technology and system based on the multi-source biological information decoding,the National High Technology Research and Development Program of China(2015AA042303),in this paper,the inherent physiological mechanism of muscle synergy is integrated into traditional motion pattern recognition.A gesture and upper limb motion pattern classification and recognition method based on the feature parameters of muscle synergy mode is designed.And through analyzing the influence of exercise mode on the synergistic characteristics of myoelectricity,explores its feasibility and robustness in gesture recognition.At the same time,explore the intrinsic link between the upper extremity reaching the target position and the electromyography synergy mode.In order to improve the recognition efficiency of the upper limb movement intention,and to adapt to the current development needs of the myoelectric prosthetic smart control strategy.In order to fully consider the physiological mechanisms inherent in limb movement during exercise pattern recognition,this paper designs a method based on Non-negative Matrix Factorization(NMF)to extract the muscle synergies to construct the feature matrix.By recording the surface electromyography(sEMG)data of the healthy participants forearm surface during the five kinds of gestures were completed,the all gestures are commonly used in daily life.The 6-channel surface EMG signal is preprocessed to extract the envelope signal.Extracting the muscle synergy pattern using NMF decomposition and constructing the eigenvalue matrix.Then,the classification effect is judged based on the distribution of the feature values in the feature space and the corresponding recognition result.The results show that the method based on muscle synergistic construction of eigenvalues has good clustering in feature space;the support vector machine(SVM)and K-means clustering classification result show that the method is feasible in gesture recognition.The relationship between the number of participants and the recognition rate by constructing the feature set proves that the method is robust.At the same time,in order to explore the motion mechanism of eigenvalues based on muscle synergy in different spatial positional motion patterns of upper limbs.This paper designs an upper limb grab motion experiment with 9 target positions in space.A total of 11 healthy participants were recruited to participate in the experiment,the surface EMG signal data during the execution of the experimental tasks of the nine spatial target positions was collected.After extracting the envelope signal by preprocessing,the NMF decomposition algorithm is used to extract the muscle synergy to construct the eigenvalue matrix.Then based on the relationship between the Variance Account For(VAF)value and the change in the number of muscle synergies,the influence of the change in the number of muscle synergies on the recognition of the upper limb spatial motion pattern is explored.The results show that there is a higher similarity between the most central target location and other locations.And with the increase of the number of synergistic elements,the difference between the edge space position and the center position recognition rate is more obvious.The result is supported by the recognition rate result and the correlation coefficient calculation result.This paper proposes a method based on NMF algorithm to extract muscle synergistic construction feature matrix and traditional motion pattern recognition.Considering the inherent physiological mechanisms of motion in motion pattern recognition,and the ideal recognition effect is obtained and the robustness is achieved.At the same time,the motion mechanism of the upper limbs in different spatial positions is explored.These are the follow-up electromyography prostheses and Rehabilitation system control provides new technical ideas.