互信息引导下的前向搜索脑—机接口导联选择算法

Forward searching channel selection algorithm based on mutual information for brain-computer interface

在脑—机接口(brain-computer interface,BCI)系统中,利用高密度导联来获取脑电信号(EEG)空间信息,增加了特征提取和识别的复杂度和难度。针对脑电信号分类识别中的导联选择问题,提出一种互信息引导下的前向搜索导联选择算法,首先根据互信息及最大相关最小冗余(maximum dependency with minimum redundancy,mRMR)原理对各个导联进行排序,以排序靠前的导联信号分类准确率为判据,采用前向搜索算法依次选择后续导联,获得最优导联组合。以BCI competitionⅣdata sets 1为分析数据集,实验结果表明,所提算法在减少导联的同时提高了BCI系统的识别率,为BCI系统的应用提供了技术参考。

In the BCI system,the use of high density leaded to obtain the spatial information of EEG signals,which increased the complexity and difficulty of feature extraction and recognition.Regarding to the channel selection problem during the classification of EEG signals,this paper proposed a novel forward searching under the guidance of mutual information-based channel selection method.Firstly,according to the mutual information and maximum dependency with minimum redundancy(mRMR)principle to sort each channel.The classification accuracy of channels on the top of list was the criterrion,and in order to the optimal channels combination,it adopted the forward search algorithm to selected the follow-up channels.Taking BCI competitionⅣdata sets 1 as the analysis of data sets,the experimental results show that the proposed algorithm improves the recognition rate of the BCI system while reducing channels and provides a technical reference for the application of BCI system.