Eye state EEG signal classification using complex valued neural classifiers

Abstract

The main source for the operation of Brain Computer Interface (BCI) technique is EEG signals. BCI is used as a direct communication between the brain and the external device. It is mainly used for supporting, augmenting, or repairing human cognitive or sensory motor function. The proposed work is the classification of EEG signals measured under eye opening and closing state using advanced learning classifiers. Eye Opening and closing (motor imagery) dataset is a benchmark data obtained from UCI (University of California Irvine). The advance learning classifiers considered are ELM (Extreme Learning Machine), PE-CELM (Phase Encoded Extreme Learning Machine and FC-FLC (Fully Complex Valued fast learning classifiers). ELM is a fast learning real valued classifier which selects input weights randomly and computes weight of the output layer analytically. PE-CELM and FC-FLC are recently developed complex valued neural classifiers and they are used in the EEG signal classification task. Generally real valued networks have less computational ability when compared to complex valued neural networks. Hence, PE-CELM and FC-FLC performs better than ELM classifier due to the orthogonal decision boundary, natural property of complex valued signal.