Efficient method for classification of alcoholic and normal EEG signals using EMD

The Journal of Engineering

关键词: skewness features;    nonstationary signals;    EMD;    entropy;    heart;    pathological activity;    nonlinear signals;    Hilbert transform;    empirical mode decomposition;    kurtosis features;    electroencephalogram signal;    radial basis function kernel;    normal EEG signal classification;    liver;    stationary intrinsic mode functions;    immune system;    negentropy;    electrical representation;    alcoholic EEG signal classification;    automatic classification method;    feature extraction;    physiological activity;    polynomial kernel;    LS-SVM classifier;    brain disorders;    b;    limited signals;    least squares support vector machine classifier;   

DOI  :  10.1049/joe.2017.0878

学科分类：工程和技术（综合）

来源: IET

【 摘 要 】

The electroencephalogram (EEG) signal is an electrical representation of brain's working that reflects various physiological and pathological activities such as alcoholism. Alcohol can affect whole parts of the body but, it particularly affects the brain, heart, liver, and the immune system; its effects on the brain are called brain disorders. Nowadays, automatic identification of alcoholic subjects based on EEG signals has become one of the challenging problems in biomedical research. In this study, an automatic classification method for classifying alcoholic and normal EEG signals, based on empirical mode decomposition (EMD), is proposed. The uniqueness of EMD method is to decompose non-stationary and non-linear signals into a set of stationary intrinsic mode functions (IMFs) that are band limited signals. These IMFs are transformed into analytic representations by applying the Hilbert transform. From these analytic IMFs, various features namely mean, kurtosis, skewness, entropy, and negentropy are extracted; these features are used as input to least squares support vector machines (LS-SVMs) classifier with radial basis function (RBF) kernel and polynomial kernel. The accuracy results achieved for LS-SVM classifier with polynomial and RBF kernels are found to be 96.67 and 97.92%, respectively, which are found to be better as compared with other state-of-the-art methods.

【 授权许可】

CC BY