Research On Recognition Method Of Lower Limb Motion Intention Based On Surface Emg Signal

With the continuous development of science and technology,exoskeleton robot technology has become a research hotspot in military science and technology,rehabilitation medicine and other fields.The advantage of exoskeleton robot is that it can respond according to the movement intention of human body and assist people to complete corresponding actions.There are many ways to obtain human motion intention,among which surface EMG signals have good real-time performance,are easy to collect,and can accurately reflect lower limb motion intention,so they are widely used.The intention of lower limb movement can be expressed by gait phase,the gait information of subjects walking on flat ground is collected as the research object,and a lower limb movement cycle is divided into multiple gait sub-phases.Different gait sub-phases correspond to different physiological activity levels of muscles.Therefore,human motion intention can be identified by extracting lower limb surface EMG signals.In this paper,the surface EMG signals of human lower limbs are collected,and the EMG signals are denoised and preprocessed,and the features are extracted,and the different gait phase patterns of human walking on flat ground are classified and identified.Firstly,the gait of human walking is analyzed,and the whole walking process is divided into several same gait cycles.In the same cycle,the gait is divided into five phases: the initial stage,the middle stage and the end stage of support and the initial stage and the end stage of swing.By analyzing the generation mechanism and biological characteristics of surface EMG signals,four main muscles in the process of lower limb movement are selected as the research objects,and the surface EMG signals of these four muscles are collected for gait phase pattern recognition.Secondly,aiming at the weakness and instability of EMG signals,wavelet threshold denoising and Butterworth filtering are used to preprocessEMG signals,which improves the signal-to-noise ratio.Using the method of moving data window,the surface EMG signal of the active segment of lower limb muscle is effectively extracted,and the signal of the resting segment is removed.Combining time domain and frequency domain features to form feature vectors,principal component analysis is used to extract the most effective features for recognition.Finally,the traditional SVM pattern recognition algorithm is introduced.Aiming at the problem of inaccurate classification of SVM algorithm near the hyperplane,KNN algorithm is integrated to optimize it,and SVM-KNN algorithm is studied to improve the classification accuracy of data near the hyperplane.Two parameter optimization algorithms,genetic algorithm and particle swarm optimization algorithm,are used to optimize the penalty function and kernel function of SVM algorithm.The experimental results show that the SVM-KNN algorithm based on particle swarm optimization has the highest accuracy for gait phase recognition and shortens the recognition time.