Pattern Recognition of EEG Based Hand Activities Using Artificial Neural Network

摘要

With the fast development of Brain Computer Interface (simply called BCI), Electroencephalography (simply called EEG) will be another interesting bio-electrical signal applied in the dexterous control of the robot hand after EMG. In order to realize it finally, pattern recognition of human hand activities based on EEG is a very important and elementary research objective. After discussing our research methodology, a two-channel measuring system about EEG signal is at first set up in this paper. Based on this system, the first experiment to locate the optimal measuring position is done to verify P3 and P4 sites are the optimal measuring positions, where the EEG electrodes are placed on the scalp in international 10/20 system for our research objective. And then the second experiment, to extract the features of hand movement and the other usual accompanying mental tasks, such as eye blink, to see the red color and listening music, is also done so as to find an important method of feature extraction, and its result shows the distribution rule of the spectrum result based on FFT analysis is a very good pattern vector for the classification of these basic mental tasks. At last, an artificial neural network classifier is presented. After sample learning is over, the artificial neural network can output good results of pattern recognition about human hand activities according to input spectral features of these mental tasks.