Ocular artifacts removal from EEG signals using Discrete Wavelet Transform and Quadratic Regression Method

Abstract

Interference of ocular artifacts (OAs) misled our perception of brain activity. In this paper, a combination of Discrete Wavelet Transform (DWT) and quadratic regression method is presented for the removal of ocular artifacts from the contaminated EEG channels. The proposed approach is using the Fp1 channel to estimate ocular artifact as it is highly contaminated by the eyeblinks and eye movements. DWT is applied on a signal from the Fp1 channel to get an estimate of ocular artifact and this noise estimate is used as a noise reference in the quadratic regression method to remove the ocular artifacts. The use of the Fp1 channel for the estimation of ocular artifact eliminates the requirement of additional electrooculogram (EOG) channels. The proposed approach is compared with the other noise removal techniques i.e. quadratic non-linear regression method, adaptive filtering (NLMS) based method, combination of Wavelet Transform, and Independent Component Analysis (wICA). The proposed approach is evaluated on a publicly available semi-simulated dataset. Three performance measures (Root Mean Squared Error, Power spectrum distortion, and Signal to Artifact ratio) are adopted to test how successfully algorithms remove ocular artifacts and how much the EEG signals are distorted after the artifact rejection procedure. Experimental results show that the proposed approach performs better than the other three methods compared. The proposed approach enhances the EEG signals and will contribute to improving the classification accuracy of BCI applications.