Multivariate iterative filtering for multichannel EEG signal processing

Abstract

The human brain is a highly complex organ that contains 100 billion neurons interacting with each other to perform day-to-day tasks. Electroencephalogram (EEG) is the recording of electrical activity of the brain results from the summations of excitatory and inhibitory postsynaptic potentials of relatively large groups of synchronously firing neurons. The processing of EEG signals has become a cornerstone in neurophysiological research and clinical diagnostics, providing insights into brain function and aiding in the development of brain-computer interface (BCI). EEG provides high temporal resolution with limited spatial resolution. A higher number of electrodes are used to record multichannel dense EEG signals for improved spatial resolution. This thesis extends the univariate adaptive signal decomposition technique, iterative filtering, to multivariate iterative filtering (MIF) for analyzing multichannel signals. Based on MIF, this thesis presents novel methodologies for the analysis of EEG signals, addressing critical challenges such as feature extraction, classification for neurological disease diagnosis, and BCI applications.