Research On Classification Of Upper Arm SEMG Signals Based On Eye-Movement Teaching

Surface Electromyography Signal(sEMG)is result from muscle contraction and it's a small signal that appears on skin.After study these signals can reflect muscle's action status to a certain extent.At present,sEMG signals have been studying deeply on rehabilitation therapy,clinical medicine and Controlled Prosthetics ECT.Before use them to control equipment such as artificial limbs,we should collect them and classify features.The feasibility and accuracy of classification depends on teaching methods.At present,teaching signals come from mainly by two methods,keyboard input and programming to set the time window.But both methods have their own shortcomings for patients with upper limb disability.So these shortcomings will impact accuracy and feasibility.This article designs an eye movement teaching system to collect and classify surface EMG signals.This article designs a system of eye movement teaching with the help of Tobii eye tracker and the board of eye movement teaching.Use Kalman filtering algorithm to optimize this system.Design some experiments to test accuracy of the eye movement teaching system.And design experiments to verify accuracy of the teaching system.Select the signal acquisition equipment of EMG-USB2+.Then acquire sEMG signals using the system of eye movement teaching.Design methods of signals acquisition and arm movement.Use the teaching signals to distinguish sEMG signals of different arm behaviors.Select the method based on wavelet analysis in time-frequency domain to extract feature values.By comparing distribution of feature values determine the feature vector.At last,design a BP neural network classifier.Use the classifier to Classify feature values with the help of eye movement teaching signal.And check the accuracy of the classifier by experimental analysis.Among all the chapters,the eye movement signal optimization algorithm,processing original sEMG signals,feature extraction and BP neural network have been simulated using Matlab simulation software.Design experiments to text the BP neural network Classifier.