Contenu principal de l'article
Adaptive noise canceler (ANC) is one of the most common methods of noise cancelation from EEG signals. However, there are two major problems with the adaptive noise cancelation method of EEG signals: 1. The reference signal is not available for the adaptive filter which should be an estimate of the contaminant noise. 2. The MSE criterion is usually used to minimize error in the adaptive filter. Since the EEG signal and EOG artifact are non-Gaussian, it is not appropriate to use the MSE criterion which only considers the second-order error. In this paper, we have used an adaptive noise removal system in which using a DWT we first derive an estimate of EOG noise and this signal is given to the ANC reference input. It also uses the error entropy criterion instead of the MSE to minimize the error signal. The simulation results show the superiority of the proposed system over other techniques in terms of RRMSE, SNR rate, and coherence analyses.