Data Acquisition And Analysis Of Electromyography Signal For Controlling Artificial Hand

Artificial hand with myoelectricity control is a new type of dynamic artificial hand and the model "man-machine system" which is controlled by biological electricity. Its principle is to utilize the change of EMG degree from the muscle of stump to be regarded as control signals, to control the movement of artificial hand, to replace the arm that people lost on the body. Compared with other control ways, it has a lot of superiority. By this reason, it is favored by patient and has wide markets, and now it has became a focus in studying of artificial limbs.The surface electromyography (SEMG) is a kind of complex result with time and space feature produced by muscle activity. It is a sort of bioelectricity signal detected without hurt and has important influence in information sending of nerve system, basic medicine research, clinic diagnosis, sport medicine and rehabilitation engineering. The research for SEMG concerning how to apply SEMG as clinic diagnosis, telemedicine and basic medicine research has became research focus in the medicine field and rehabilitation field.Two problems about the real-time collection of SEMG and model classification have been studied in this paper. At the same time real-time collection system has been realized and three typical action models have been recognized and classified. The main research work and its production are as follows:(1) Based on the characteristic of SEMG, a real-time collection system has been realized and the signal data has been saved.(2) The SEMG signal that collected are analyzed in the area of time, amplitude and frequency, thus the characteristic of SEMG is found out. The analyzed experimentation signals show that the analyzed in the area of frequency is very efficiency and three typical action models have been recognized and classified. The analyzing of power spectrum can prove the feasibility of frequency band energy decomposition using wavelet packet. So it proved in theory that it is feasible to identifying the characteristic of SEMG and recognizing different typical action. (3) In order to use much more feature extraction methods and much accurate, neural network are used. Frequency band energy decomposition using wavelet packet input into the artificial neural network and then three typical actions have been classified. All the soft wares have been programmed with MATLAB and the inspiring results have been acquired at last.