基于收缩LDA的ERP成分分类研究

LDA是重要的分类方法,应用LDA于ERP成分分类时需要对高维空间协方差矩阵进行准确的估计,这个估计是一个复杂且难以求解的问题.为了改善这个问题,提出使用收缩估计,并证明了应用收缩估计的LDA产生的分类结果远远优于原始的LDA.

Classification of ERP Components Based on LDA with Shrinkage

LDA is one of the most important classification techniques. However, the bottleneck of the technique requires an accurate covariance matrix estimation in high dimensional sensor spaces which is a highly intricate problem. As a remedy, using shrinkage estimators is proposed and the appropriate regularization of linear discriminant analysis （LDA） by shrinkage yields excellent results are shown.