Research Of Surface EMG Signal Mode Classification Method Of Arm Action Based On SVM

Arm action Arm action surface electromyogram signal uses electrical signalsproduced by muscles which is recorded by surface electrode under arm skin surface, itcan quantify functional status of nerve and muscle when arm is acting. As theextraction mode of the surface electromyogram signal has the advantages ofconvenient,accurate and non-invasive, it is widely studied and applied inrehabilitation medicine and sport medicine and intelligent robot. With thedevelopment of information science and technology, accurately selecting effectivecharacteristics from surface electromyogram signal and achieving mode patternrecognition in high resolution according to the signal characteristics has become thekey of controlling bionic prosthetic by electromyogram signal. This paper issupported by Jilin province technology development project research and develop ofbionic arm with temperature and tactile telepresence (NO:20090350), implementsresearch on arm action surface electromyogram pattern classification method, so as topromote the practicability of controlling bionic prosthetic by electromyogram signal,this paper has important scientific value and social significance.The main work of this paper is:(1) Based on the characteristics of the suface electromyogram signal, combinedwith local anatomy knowledge, make clear the two largest contribution muscles whenarm is acting, ascertain the vantage point when the electrode is extracting signals, usepatch electrode and electromyogram signal acquisition instrument to perform thework of extracting surface electromyogram signal of arm common action pattern.(2) using time domain analysis method, frequency domain analysis method andtime frequency analysis method to perform feature extraction of the surfaceelectromyogram signal which is extracted, by analyzing data results, it comes to aconclusion that time domain analysis method and frequency domain analysis methodhas its one-sidedness, ascertain using wavelet packet method based on time frequencyanalysis to extract the feature of surface electromyogram signal, ultimately using thevariance and energy of the wavelet packet coefficient as the feature vector elementswhich form feature vector.(3) analyse the main method of pattern recognition, use standard support vectormachine, least squares support vector machine and BP neural network algorithm to perform pattern recognition of the feature vector. Carry out the discussion andexperiment of parameter selection method, analyze the characteristics of the threekinds of parameter optimization methods.(4) Carry out statistical analysis of the correct recognition rate and training timeof the three pattern recognition algorithm, ascertain the advantages of using particleswarm optimization method to perform parameter optimization and using leastsquares support vector machine to perform pattern recognition as well, it has higherrecognition rate and shorter operation time. Although parameter optimization haslonger operation time, the training and testing of the classifier model can be dividedinto two steps, by drawing parameter optimization into classifier model training step,the time of using acquired classifier model to perform action pattern recognition canbe greatly shortened.(5) Use Matlab GUI module development arm movement pattern recognitionsystem offline, the various aspects of surface EMG signal processing integration, thesystem becomes visual, easy to operate.