An investigation of EEG artifacts elimination using a neural network with non-recursive 2nd order volterra filters

Abstract

The artifacts caused by various factors, EOG (electrooculogram), blink and EMG (electromyogram), in EEG (electroencephalogram) signals increase the difficulty in analyzing them. In addition, EEG signals containing artifacts often cannot be used in analyzing them. So, it is useful and indispensable to eliminate the artifacts from EEG signals. A neural network with non-recursive 2nd order volterra filters is used to eliminate the artifacts from EEG signals. The proposed method is a new approach in respect to slotting a non-recursive 2nd order volterra filter into individual neurons of a neural network. First of all, in order to investigate the usefulness of the proposed method in eliminating the artifacts from EEG signals, we apply it to the artificial EEG signals mat are weakly stationary process. As the result, the artifacts can be eliminated from EEG signals almost exactly using the proposed method, and ft is suggested the proposed method should be useful in eliminating the artifacts from EEG signals.