Estimation of evoked potentials using high order statistics-based adaptive filter

This paper is to present a high order statistics-based adaptive interference cancel filter (AIC-HOS) to process evoked potential (EP). In the conventional ensemble averaging method, experiments have to conduct repetitively to record the required data. In normalized LMS adaptive filter, inappropriate step size always causes deficiency. This AIC-HOS system has none of the above disadvantages. This system was experimented in somatosensory evoked potential corrupted with EEG. A gradient type algorithm is used in this AIC-HOS structure to regulate the SNR of EEG and EP. This method is also simulated with visual evoked potential and audio evoked potential. The results obtained are satisfactory and acceptable in clinical usage. The AIC-HOS is superior to normalized LMS using adaptive filter in that it converges easily. Moreover, it is not sensitive to selection of step size in stabilities in convergency.