近似熵与SVM在自动分类癫痫脑电信号中的应用

The Application of Approximate Entropy and Support Vector Machine in Classifying Signal of Epilepsy

采用近似熵(ApEn)与支持向量机(SVM)相结合的方法对癫痫发作间歇期EEG和发作期EEG进行自动识别分类,通过分类结果来检验非线性动力学指标是否可有效运用于脑电癫痫波的自动实时探测中并验证由非线性动力学指标训练的分类器的泛化能力。研究使用参考患者的脑电数据构建SVM分类器,在此分类器的基础上,对其他患者的脑电数据进行分类得到结果。结果表明利用非线性动力学指标训练的分类器具有良好的泛化能力,对不同患者的脑电数据的平均分类准确率达到较高水平。

To classify the EEG signals into interictal EEGs and ictal EEGs by the method of combination with ApEn and SVM,examine whether the nonlinear dynamic index can be effectively used in automatic detection of EEG epilepsy wave and the generalize ability of classifier trained by non-linear dynamics through the classification result.We used EEG from epileptic patient to train SVM and used it to classify EEG from other epileptic patients.It show that the SVM classifier practiced by nonlinear dynamic characteristics has a good generalizing ability;the classifier achieves a good classification result to different epileptic patients.