EEG Signal Separation using Improved EEMD - Fast IVA Algorithm

Abstract

Blind Source Separation (BSS) is one of the promising approaches to retrieve the information from the non Gaussian independent components of the mixtures. The number of sources and the mixing method of the sources are unknown and hence the term “blind”. Joint blind source separation (JBSS) is a means to extract common sources simultaneously found across multiple datasets, e.g., electroencephalogram (EEG). In this study, by separating the signal into its possible independent components, the simplification and comprehension of analysis of EEG signals was aimed. In this paper, Improved EEMD-Fast IVA algorithm is proposed to separate into its possible common independent vector found across multiple channels. The investigation on simulated EEG signals demonstrates the better result of the proposed algorithm. Moreover, a comparative study of Improved EEMD-Fast IVA with the reported BSS methods like STFT-ICA, Wavelet-ICA and IVA is presented. Through such an analysis, it was thought that early diagnosis of any neurological disease such as epilepsy, Parkinson's disease, sleep disorders as well as information regarding the location and size of problematic zone become possible.