基于遗传蜂群算法的运动想象BCI系统导联选择*

在运动想象脑-机接口系统中,常采用高密度导联获取脑电信号,导致实验准备时间长,系统运行速度慢,性能变差等问题。针对上述不足,本文提出了一种基于遗传算子的蜂群算法用于导联优选,引入交叉和变异算子以提高蜂群算法的邻域搜索能力。通过对第四届国际BCI 竞赛 Dataset 1中四名被试者(a,b,f和g)的59导联运动想象数据进行导联优选,用多类CSP算法和支持向量机对优选导联数据进行特征提取和分类识别。结果表明所提出的算法在大大降低了导联维数的同时,也得到了比采用全部导联更高的分类识别率,验证了本文所提算法的实用性和有效性。

Channel selection for motor imagery brain-computer interfaces based on generic bee colony algorithm

In motor imagery brain-computer interfaces systems,high-density multi-channels are often used to acquire electroencephalogram (EEG) however excessive channels can lead to some problems, such as long preparation time, slow running speed and poor performance. To address the questions, this paper proposed an improved artificial bee colony algorithm based on genetic operators (GA-ABC) to select the best channels. Crossover and mutation operators were introduced to improve the neighborhood search ability of the bee colony algorithm and to avoid the local optimization of the bee colony algorithm. The proposed algorithm was used for the data sets of the subjects a, b, f and g of BCI competition IV Dataset 1, whose channel number is 59. Multiclass CSP and the support vector machine (SVM) were used to extract and classify the features of EEG, and then this paper compared the classification accuracies of the whole channels and the selected channels. The results show that the proposed algorithm can greatly reduce the number of channels and lead to a higher accuracySthan the full channels, which verifies the practicability and effectiveness of the proposed algorithm.