Comparison on Performance of Adaptive Algorithms for Eye Blinks Removal in Electroencephalogram

Abstract

The interference of eye blink artifacts can cause serious distortion to electroencephalogram (EEG) which could bias the signal interpretation and reduce the classification accuracy in a brain-computer interface (BCI) application. To overcome this problem, an algorithm to automatically detect and remove the artifacts from EEG signals is highly desirable. One of the methods that can be applied for automatic artifacts removal is adaptive filtering through an adaptive noise cancellation (ANC) system. In this paper, we compare the performance of three adaptive algorithms; namely LMS, RLS, and ANFIS, in removing the eye blink from EEG signals. To evaluate the results, the SNR, MSE and correlation coefficient values are calculated based on the results obtained by using one of the widely used methods for blinks removal, independent component analysis (ICA). The results show that RLS algorithm provides the best performance when comparing with the ICA method.