Classification of Electroencephalography signals using mixture of Features

Abstract

Electroencephalography (EEG) signals provide valuable information to study the brain function and neurobiological disorders. Digital signal processing gives the important tools for the analysis of EEG signals. The primarily focus on classification of EEG signals using different feature extraction methods for pattern recognition purpose. The various tools are used for extracting the relevant information from EEG data is Discrete Wavelet Transform (DWT), Spectral analysis using Autoregressive (AR) Model and Lyapunov Exponents. The EEG data was collected from standard repository source. The two classifiers ANN and CNN are used for the classification purpose. A technique is proposed based on using the combined features extracted from different methods. In committed neural network, several independent neural networks are trained by the extracted features from different EEG signals are constituted a committee. This committee takes the final decisions for classification which in turn represents a combined response of the individual networks. The performance of the proposed algorithm is evaluated on 300 different recordings from three different cases comprising of healthy volunteers with eyes open, epilepsy patients in the epileptogenic zone during a seizure-free interval, and epilepsy patients during epileptic seizures. The experimental results show that the classification performance for the proposed technique is higher than some of the earlier established techniques.