基于优化核参数支持向量机的意识任务分类

CLASSIFICATIONS OF EEG DURING MENTAL TASK BASED ON SUPPORT VECTOR MACHINE WITH OPTIMAL KERNEL-PARAMETER

根据支持向量机的基本原理,给出一种推广误差上界估计判据,并利用该判据进行最优核参数的自动选取。对三种不同意识任务的脑电信号进行多变量自回归模型参数估计,作为意识任务的特征向量,利用支持向量机进行训练和分类测试。分类结果表明,优化核参数的支持向量机分类器取得了最佳的分类效果,分类正确率明显高于径向基函数神经网络。

The fundamental of support vector machine (SVM) based on structure risk minimization was introduced. An estimation formula of upper bound of generalization error was given, and the optimal kernel-parameter of the SVM was selected automatically by the formula. The feature vectors were extract-ed from six-channel electroencephalograph (EEG) data segments of four subjects under three mental tasks by the mean of a multivariate autoregressive (MVAR) model method. These vectors were considered as the inputs of classifiers to test classification accuracies for three task pairs. Average classification accura-cies indicated that the optimal kernel-parameter method could get optimal results, and was significantly better than that of Radial Basis Function (RBF) network.