An approach for pattern recognition of hand activities based on EEG and Fuzzy neural network.pdf

Journal of Mechanical Science and Technology, Vol. 19, No. 1, pp. 87-- 96, 2005 87 An Approach for Pattern Recognition of Hand Activities Based on EEG and Fuzzy Neural Network Xiao Dong Zhan School of Mechanical Engineering, Xian Jiaotong University, Xian, Shaanxi Province 710049, China Taehun Kang, H y o u k Ryeol Choi* School of Mechanical Engineering, Sungkyunkwan University, 300 Chunchun-Dong, Jangan-Gu, Suwon 440-746, Korea Electroencephalography (EEG) is another interesting bio-electrical signal to differ from EMG (Electromyography). In order to pursue its application in the control of the multi -fingered robot hand or the prosthetic hand, the pattern recognition technology of the human hand activities based on EEG should be investigated as a very important and elementary research objective at first. After discussing our research strategy about EEG applied in the control of the robot hand, the recognition model named as Fuzzy Neural Network (FNN) is set up in this paper, and then its related algorithms, such as the fundamental knowledge produced, the learning samples set, the features extracted, and the patterns recognized with the artificial neural network (ANN), are deeply discussed for achieving the classification of some basic mental tasks. In addition, the experimental research has also been done using a two- channel system of measuring EEG signal, and the result shows the new recognition model using FNN can extract not only the effective spectral features of the hand movements and the other usual accompanying mental tasks, such as blinking eyes, watching red color and listening music, so as to achieve the fundamental knowledge production and the feature extraction, but also has the good capability of the pattern recognition about the human hand activities through the fuzzy setting of the learning samples and the training of its ANN. Key Words : Electroencephalography, Pattern recognition, Artificial Neural Network 1. Intro