Bi-spectral higher order statistics and time-frequency domain features for arithmetic task classification from EEG signals

Abstract

Recently the development of brain-computer interface applications has drawn the attention of researchers as it can assist physically challenged people to communicate with their brain electroencephalogram signal. In this paper, a Brain-Computer Interface (BCI) is designed using EEG signals to differentiate between two mental arithmetic tasks performed by the same subject. The presented BCI approach includes three stages: (a) Elimination of power-line frequency components and segmentation of the raw signals (b) bi-spectral analysis as well as time and frequency domain features extraction (c) classification using SVM classifier. Bi-spectrum is proposed in order to characterize the non-Gaussian information contained within the EEG signals. Higher-order statistics of Bi-spectrum are used for classification. Time domain features as Generalized HFD and frequency domain features as frequency band powers and asymmetry among the channels are used to enhance the classification performances up to 100