Eeg Signal Classification Method Based On Neural Networks

Electroencephalogram (EEG) signal is an important information source to study the brain activity, and the communications between people and computer based on the EEG becomes a new way of man-machine interface. It has been discovered that some physiological electricity signals will appear in the brain when one person wants to do something. These homologous physiological signals reflect his or her intention and come true it by controling the muscles. This paper first introduces EEG research background and development of the status, descript briefly acquisition of EEG singal and then discuss the current analysis of EEG in some popular methods. The paper focused on discussing EEG classification methods. Finally, this paper introduces the experiment. in particular based on Neural Networks 2-5 kinds of mental state of the EEG classification methods, through time-domain regression method for EEG Denoising pretreatment, and then uses of 6-order AR parameter model extracted EEG as a neural network input, with psychological operations of diffirent thinking brain electricity siginals.and finally with Matlab 7.0 simulation, experiments show that this method can achieve good classification results.This paper mainly studies and discusses on the following questions: It has done noise reduction to the EEG through the time-domain regression method.then disparted sections on the treated EEG signals, and then extracted features to the former and latter EEG signals with AR parameters, then classified with BP neural network and PNN neural network.Finally, comparing the classification results.And tested by BCI II of the 2002 data, the experiments have reached very good classification effect.