Analysis and classification of electroencephalography signals

Abstract

EEG signal processing is one of the hottest areas of research in digital signal processing applications and biomedical research. Analysis of EEG signals provides a crucial tool for diagnosis of neurobiological diseases. The problem of EEG signal classification into healthy and pathological cases is primarily a pattern recognition problem using extracted features. Many methods of feature extraction have been applied to extract the relevant characteristics from a given EEG data. The EEG data was collected from a publicly available source. Three types of cases were classified viz. signals recorded from healthy volunteers having their eyes open, epilepsy patients in the epileptogenic zone during a seizure-free interval, and epilepsy patients during epileptic seizures. The feature extraction was done by computing the discrete wavelet transform and spectral analysis using AR model. The wavelet transform coefficients compress the number of data points into few features. Various statistics were used to further reduce the dimensionality. The AR coefficients obtained from burg auto-regressive method provide important features of the EEG signals. Classification of the EEG data using committee neural network provides robust and improved performance over individual members of the committee. F-ratio based dimension reduction technique was used to reduce the number of features without affecting the accuracy much.