基于小波包分解的脑电信号特征提取

EEG feature extraction in brain computer interface based on wavelet packet decomposition

在脑机接口研究中,针对脑电信号的特征抽取,提出一种基于小波包分解的方法,利用Fisher距离准则,选择具有较大可分离性的特定子带小波包系数和能量作为有效特征,构成特征矢量,并采用BCI2003竞赛数据,通过对该特征矢量的可分性和识别精度2个指标的评估,表明了所提出方法的有效性。

In the study of brain computer interface, a method based on wavelet packet decomposition is proposed which is used for the feature extraction of electroencephalogram. The power of special sub-bands and coefficients of wavelet packet decomposition that have larger separability are selected to construct eigenvector according to Fisher distance criterion. The eigenvector is obtained by combining the effective features of electroencephalograph signals from different channels. The performance of the eigenvector is evaluated by separability and recognition accuracy using the data set from BCI 2003 competition. Classification results have proved the effectiveness of this method.