An Adaptive Noise Cancellation System Based on Linear and Widely Linear Complex Valued Least Mean Square Algorithms for Removing Electrooculography Artifacts from Electroencephalography Signals

In this study, an adaptive noise cancellation (ANC) system based on linear and widely linear (WL) complex valued least mean square (LMS) algorithms is designed for removing electrooculography (EOG) artifacts from electroencephalography (EEG) signals. The real valued EOG and EEG signals (Fp1 and Fp2) given in dataset are primarily expressed as a complex valued signal in the complex domain. Then, using the proposed ANC system, the EOG artifacts are eliminated in the complex domain from the EEG signals. Expression of these signals in the complex domain allows us to remove EOG artifacts from two EEG channels simultaneously. Moreover, in this study, it has been shown that the complex valued EEG signal exhibits noncircular behavior, and in the case, the WL-CLMS algorithm enhances the performance of the ANC system compared to real-valued LMS and CLMS algorithms. Simulation results support the proposed approach