Classification Algorithm Based On The Movement Of The Surface Emg Studies

With the rapid development of science and technology, substantial changes for people's life are being or will be changed by combining the electric technology and the Bioinformatics. It has become the researchers'focus that people probe the mystery of life with the signal processing's tools. In this paper, the modern signal processing and the pattern recognition technology are used to classify eight actions, based on the Research Programs of Science and Technology Commission Foundation of Yunnan Province.Based on the EMG, the meaning of the action recognition in the production and life will be analyzed at first in this paper. Then the production and transmission mechanism for EMG signal of the Skeletal Muscle as well as the non-stationary of the EMG will be analyzed and summarized. Besides, a circuit for acquiring and amplifying the surface EMG signal has been designed while the Digital Acquisition System (DAS) based on the embedded system and USB as well as the PC software which services for DAS have been developed before the research. The aim signals with length of256points at the1K Hz sample frequency were chopped into eight segments. The240characteristics consisted of mean absolute value, mean absolute value, zeros crossings, waveform length, coefficients, AR model and wavelet were extracted section by section. The existence of the laws in the EMG signal was proofed by the correlation coefficient. At last, the start time of the actions in the EMG signals which is very important in the actions recognition was found. The characteristics for80samples composed10actions were extracted. The dimension of the eigenspace was declined by applying PCA as the result was classified by the ANN BP Networks. And the recognition rate is96%.