Linearity Abstractions in Scene Perception: Evaluating EEMD for Maximizing Task-related Information in Event Related Potentials

Linearity extraction is complex higher level cognitive process. It involves non-linear and transient operations which can be best captured with signal analysis methods that assume non-linearity and non-stationarity in the data. In the present paper, we evaluate Ensemble Empirical Mode Decomposition (EEMD) in ERP data recorded during linearity abstraction task. EEMD as a datadriven denoising process, has a high signal retention percentage when compared to FIR denoising. On low SNR datasets, it shows fairly high degree of noise suppression as given by Noise Suppression Index (NSI). Time-frequency analysis of the EEMD denoised ERP data shows fairly high degree of signal retention.