Single-Trial Event Related Potentials Extraction by Using Independent Component Analysis

Estimation of single-trial event related potentials (ERPs) based on independent component analysis (ICA) is investigated in this paper. The performances of four ICA algorithms (AMUSE, SOBI, Infomax, and JADE) are assessed by separating the ERPs from semi-simulated EEG datasets, in order to determine the optimal method. Infomax and JADE performs significantly better than AMUSE and SOBI in eliminating artifacts and Infomax is further successful in extracting ERPs from real EEG recordings. The excellent performance of Infomax with real case supports this ICA algorithm as a valid choice to minimize the influence of artifacts on ERPs single trial extraction.