针对基于EEG的脑—机接口（BCI）实验数据分布不明朗的特点，双滤波模式（DFP）算法利用样本模式相似性来优化BCI的分类特征———运动相关电位（MRPs）特征的空间（即电极位置）和时间投影方向，使得映射后异类样本模式差异性与同类相似性的比值最大化。该算法考虑MRPs特征对时间、空间的敏感性，并以自适应的方式挖掘它们适合分类的信息；优化时不需要进行样本数据分布假设，符合BCI数据特点。最后，DFP算法对BCI competitionⅠ、Ⅱ两组数据进行实验，识别效果均高于相关比赛的最好成绩，这表明DFP算法能有效提取MRPs特征。%Focusing on the unclear data distribution in the brain-computer interface (BCI),this paper proposed a novel ap-proach named double-filters pattern (DFP).It used the similarity to optimize the mapping directions of space (that was,elec-trode placement)and time of the movement related potentials (MRPs),which was a common classification feature in BCI.It maximized the difference in different class and the similarity in the same class.DFP took account of the MRPs’sensibility in space and time and extracted the classified information adaptively.It didn’t make any assumption about the latent data distribu-tion and was suitable to the data characteristics in BCI.At last,it applied DFP to two datasets from BCI competitionⅠandⅡ. The accuracies were better than the best results in the competitions.It shows that DFP can effectively extract MRPs features.
展开▼
