A Comparison of EEG Preprocessing Methods using Time Delay Neural Networks

Abstract

Multichannel recordings of EEG data during various mental tasks are processed using two popular methods, independent component analysis (ICA) and matching pursuit (MP). The results are fed to a time delay neural network (TDNN) for classification of each mental task. Based on the results of the test sets, we analyzed the effectiveness of ICA and MP methods for use in EEG preprocessing and TDNN classification. It is shown that ICA is more effective than MP in lowering the neural network classification error; however this advantage is not significant