基于CSSD和SVM的抑郁症脑电信号分类

从EEG脑电信号中提取与疾病相关的信息以实现对抑郁症的自动诊断。首先采用共空域子空间分解(CSSD)方法,对躁狂型抑郁症患者与健康人两组的16导联脑电信号进行特征提取,然后用支持向量机(SVM)分类器进行训练和分类测试。实验结果表明,相对于用小波变换提取的频率相关参数为分类特征的分类准确率为88%,采用CSSD方法提取特征参数进行分类可以取得更理想的效果为95%,后者的16导联脑电信号在空间模型上表现出较高的模式可分性。该研究成果对精神抑郁症的物理诊断和研究提供了新的视角。

Application of CSSD and SVM for EEG Signal Classification

It has been an important issue to get information related to the disease from EEG for auto diagnosis. This paper used Common Spatial Subspace Decomposition (CSSD) to extract features from 16-channel's Electroencephalograph (EEG) of melancholia patients and normal healthy persons respectively. Then a classifier based on Support Vector Machines (SVM) was designed for classification. Compared to the frequency features extracted by wavelet method with the accuracy of 88%, the CSSD method gave better accuracy (95%) in EEG classification. The melancholia could be identified from the tested people through 16-channel's EEG. The result could be applied to the melancholia diagnosis and clinical research.