Wavelet based single trial Event Related Potential extraction in very low SNR conditions

Abstract

Event Related Potentials (ERPs) are generated in ongoing brain electrical activity due to visual, auditory, or sensory stimuli. These signals have very low SNR and are contaminated by background EEG. Extraction of single trial ERPs from background EEG is a challenging task due to overlapping nature of the frequency bands of ERP and EEG signals and much higher power of EEG than ERPs. In this paper we proposed a method based on wavelet transform and adaptive noise cancelers in order to extract single trial ERPs from background EEG in very low SNR conditions. Simulation results show the superiority of the proposed algorithm over the existing methods. In addition, performance of the algorithm is justified under different noise models namely White Gaussian Noise, Auto Regressive, and Real EEG signals.