基于小波变换和时域能量熵的P300特征提取算法

P300 feature extraction algorithm based on wavelet transform and temporal energy entropy

针对P300信号特征提取和分类过程中训练及测试速度相对较慢的不足,提出了一种基于P300带内带外特征的脑电信号特征提取方法,将时域能量熵和离散小波变换相结合,克服了P300信号识别中对电极数量和脑电信号叠加次数的苛刻要求。试验采用支持向量机作为分类器,在BCI Competition 2003和BCI Competition 2005的P300试验数据集上进行验证,结果表明,提出的方法只需对一导数据进行处理,只有2次叠加平均,就能得到很好的分类效果及较短的分类系统运算时间。

Aiming at the drawback of slow training and testing speed in P300 feature extraction and classification,a new P300 feature extraction method based on P300 in band and out of band EEG signal characteristics was proposed,which combines wavelet transform with temporal energy entropy.The proposed method overcomes the harsh requirement for electrode number and number of averaging.In this paper,support vector machine is used as classifier and the P300 data sets of BCI Competition 2003 and BCI Competition 2005 are used to verify the method.Verification results show that the proposed method uses only one electrode data and 2 times signal averaging,and still can obtain very good classification result and consume short computation time.