An EEG Based Approach for Pattern Recognition of Precise Hand Activities with Data Fusion Technology

Abstract

Currently electroencephalogram (EEG) is one of focused research topics as driving source of prosthetic hand and rehabilitation robot. In terms of features of EEG, the research strategy on the EEG based pattern recognition of precise hand activities is put up using data fusion technology in the thesis. In the method, after WT decomposition of original EEG for precise hand activities, power spectral density and average wavelet modulus of EEG signal in different scale space are extracted as feature vector, which are input into data fusion center based on BPNN to recognize four precise hand activities modes such as arm waiting state, arm moving, hand stretch and hand grasp. Experimental results indicates that a maximal off-line correct recognition rate of 90% and an on-line correct recognition rate of 75% are achieved for EEG of precise hand activities.