基于小波变换和神经网络的脑电信号分类方法

结合小波变换和神经网络对酒精中毒者和正常清醒者的脑电信号进行分类.通过分析脑电数据找出分类特征;采用一维离散小波变换提取含有分类特征的脑电信号频段,并以小波变换分解系数作为信号特征,实现数据序列长度压缩;对应3种刺激方式建立3个相同结构的学习向量量化(LVQ)神经网络,用于对脑电信号的预分类;根据判决规则得到最终分类结果.对真实脑电数据的分类正确率达到89%.

Method of EEG signals classification based on wavelet transform and neural networks

Electroencephalography(EEG) signals of alcoholic subjects and control subjects were classified by combination of wavelet transforms and neural networks.Classification features were discovered through the EEG data analysis.The frequency bands of EEG signals including classification features were extracted by 1-D wavelet transforms.The decomposed coefficients of wavelet transforms were remained as signals characters to accomplish the length compression of data sequences.Three learning vector quantization(LVQ) networks with same structure corresponding to three kinds of stimulations were built for the predictive classification of the EEG signals.The final classification results were acquired by judge rules.The classification accuracy of experiment EEG signals reach 89%.