Abstract Ocular artifacts are a major source of contamination in electroencephalogram (EEG) signals, thereby reducing the quality of information. Artifact removal methods play a vital role in the proper interpretation and analysis of actual brain information. This paper emphasizes the importance of identifying artifacts before their removal to preserve neural information by proposing two advanced source decomposition-based models, namely empirical mode decomposition (EMD) and empirical wavelet transform (EWT). In this study, both EWT and EMD were utilized for artifact identification and to estimate the reference artifact signals. Identified artifacts were removed using a normalized least mean square (NLMS) based adaptive filtering (AF) technique. To test and compare the efficacy of the developed models, an open source EEGdenoiseNet dataset was utilized in this study. The results obtained suggest that the empirical wavelet transform and adaptive filter-based model performed better, with an average improvement in signal-to-noise ratio (SNR) of 9.21dB and an average correlation coefficient (CC) value of 0.836734. The proposed models were further validated on real EEG data from the BCI Competition 2008 Graz dataset A, where EWT-AF achieved higher SNR compared to EMD-AF. The proposed work also aligns with Sustainable Development Goal (SDG) 3.
Raj et al. (Thu,) studied this question.