【摘要】 由于人脑对事件响应频带各不相同,为了准确确定个体最优滤波频带,提出一种多策略变异算子和时变非线性交叉因子差分进化算法对运动想象EEG频带进行处理,采用共空间模式算法提取特征向量,利用线性分类器进行分类识别。使用该方案对BCI competition Ⅲ-dataset 4a受试者的EEG数据进行了10次5倍交叉分类实验。实验结果表明,该算法稳定性强、耗时少,解决了运动想象BCI特征提取中的最优频带选择问题。更多还原

【Abstract】 Due to the fact that different people’s brain response to events is in different frequency bands,a method of multi-strategy mutation operator and time-varying nonlinear crossover factor differential evolution is proposed for accurate determination of the personal optimal filtering frequency band,and feature vectors are extracted by common spatial pattern algorithm and classified by the linear classifier. Based on this strategy,10 times 5-fold cross validation experiments for BCI competition Ⅲ-dataset 4 a EEG data of five subjects are implemented. Experimental results show that the algorithm has the advantages of strong stability,less time-consuming and strong real-time performance,and the problem of optimal bands selection in motor imagery BCI feature extraction is therefore solved.更多还原