A Research Of Pattern Recognition Based On Electroencephalogram Rhythm

In medical rehabilitation field, researchers are facing a new problem, that is how to provide effective help for the disabled people who are lack of move-ment function but have normal thought as healthy people. The study of the performance of brain-computer interface(BCI) is of far-reaching significance for alleviating the disables'pains and improving their life quality.This paper uses the competition data as the experimental data. In this paper, the author carries through the research of feature extraction and pattern classification algorithms based on the experiments of imagination of left or right hand movement and imagination of left hand or foot movement using MATLAB simulation. In this paper, the following aspects of work based on electroencephalogram(EEG) are mainly carried out:(1) Expound independent component analysis(ICA) algorithm for EEG pre-treatment.(2) Study the wavelet packet decomposition algorithm using the db4 wavelet packet. By using this algorithm, the 5 basic types of EEG rhythms, that is, the mu rhythm and alpha, beta, delta, theta rhythm, are successfully separated out.(3) Extract the features by using these algorithms which are described as followings, power spectrum estimation on the basis of modified periodogram and auto-regressive(AR) models, wavelet coefficients and wavelet entropy, and common spatial subspace decomposition(CSSD).(4) Carry out the classification based on the features extracted by using these algorithms which are described as followings, fisher linear discriminant, per-ceptron algorithm and support vector machine (SVM). Features are different, categories are different.(5) Compare the combinations of different pattern recognitions as above. By comparison, without ICA, the combination algorithm of CSSD and SVM, its accuracy can reach the highest, 74 percent accuracy. the pattern recognition algorithms described in this paper can improve the classification accuracy of BCI systems, and they can make the BCI systems obtain better working per-formance. Though this will prepare ready-made for establishing a really useful and easy controlled BCI system.