基于运动想象的脑电信号特征提取研究

Feature extraction of EEG signals based on motor imagery

基于运动想象脑电信号的脑-机接口系统在医疗领域具有广阔的应用前景,被应用于运动障碍人士的辅助控制以及脑卒的预后康复。由于运动想象的脑电信号信噪比低、不平稳以及差异性显著,对脑电信号识别带来负面影响。一个有效的特征提取算法能够提高脑-机系统的脑电信号识别率。提出一种多通道的脑电信号特征提取方法,将数据矩阵分解为基矩阵与系数矩阵的乘积,以类间离散度做为性能判据对系数矩阵进行特征提取,提取可分性更高、维数更少的特征。结合脑电信号识别领域常见的分类器在2008年BCI竞赛数据集上进行验证,证明所提方法是有效的。

The brain-computer interface(BCI)system based on motor imagery(MI)electroencephalogram(EEG)has a broad application prospect in the medical field,which can be applied to the auxiliary control of the disabled and the prognosis and rehabilitation of the brain.Because of the low SNR,instability and significant difference of EEG signal in motion imagination,it has a negative effect on EEG signal recognition.An effective feature extraction method can enhance the accuracy of EEG in BCI system.In this paper,a multi-channel feature extraction method for EEG signals is proposed.First of all,the data matrix is decomposed into the product of the basis matrix and the coefficient matrix.Then the coefficient matrix is extracted by using the inter-class dispersion as the performance criterion to extract the features with higher separability and less dimension.The experiment of BCI 2008 competition data set shows that the method is effective.