Research For Pattern Recognition Of The Bionic Arm’s Movement Based On Surface Electromyographic Signal

The surface electromyography signal has been widely used in sports medicine,clinical diagnosis and rehabilitation medicine. At present, the number of physicallydisabled persons in China is more than2412million people. Combine the field ofengineering methods and theories which are used in the development of Chinesedisabled and the elderly welfare, complying with the demand of China’s aging societyand people with disabilities has a positive significance. Currently, the bionic armshave many drawbacks such as low recognition rate and single movement pattern.Based on the above shortcomings, this paper proposed a new bionic arm movementpattern recognition method, the extracted surface EMG analysis and processing,thereby increasing the freedom of the bionic arm to make it closer to the truth of thehuman body activity.This paper first introduces the research background and significance, anddomestic and foreign literature were studied and compared to understand thedevelopment of bionic prostheses at home and abroad. Then a brief introductiondescribes the principle of the EMG signals, and some statistical mathematical modeland EMG characteristics, which can be used as a basic theory for the design of systemhardware acquisition circuit. In order to meet the needs of real-time, domain analysisof EMG signals, the general amplitude analysis, combined with a zero number and thesignal length of the two statistical characteristics, compared to multi-channelcharacteristics of the EMG signal arm Action recognition rate. In this paper, dynamiclinear model to simulate the EMG signal can be seen that the fit of the model is verygood through really signals and simulated signals. In addition, we analyzed thedynamic model of order N (1-50bands); large amounts of data obtained a minimum, a20-order model RMS error of about7.8%.Then we also introduced the time-frequency domain characteristic parameterextraction. Due to limitations of time domain and frequency domain feature extraction,we need to introduce wavelet analysis to the analysis of time-frequency domaincharacteristics. The main applications are the short-time Fourier transform, wavelettransform and wavelet packet transform. But taking into account the practical application of wavelet packet analysis can increase the freedom and precision of armmotion recognition, but because of time-frequency domain analysis of a long time,real-time requirements are not met.Finally, the arm movement pattern recognition methods, artificial neural network,support vector machines, Bayesian classifier method. Discussed the advantages anddisadvantages of each method and support vector machine classification experimentsof multi-degree of freedom arm movement.In this paper, the design of the bionic arm movement pattern recognition system,low-cost, real-time, the small RMS error and a high degree of imitation of human armelbow movement.