Automated Ocular Artifacts Removal Framework Based on Adaptive Chirp Mode Decomposition

The removal of ocular artifacts (OAs) from electroencephalogram (EEG) signals is crucial for effective and accurate analysis of the signals in different neurological and brain computer interface (BCI) applications. Towards this aspect, various multi-channel and single-channel techniques have been proposed for the removal/suppression of OAs from EEG signals. Recently, due to the rising demand of resource-limited ambulatory systems, single-channel techniques are more preferred for OAs removal. Although the existing techniques effectively remove OAs, they introduce distortion in the denoised EEG signal and the corresponding neurological features. Therefore, in this paper, we propose a framework for automated detection and removal of OAs from single-channel EEG signal based on discrete Fourier transform (DFT) and adaptive chirp mode decomposition (ACMD). The proposed framework is built in four stages: application of DFT to single-channel EEG signal; rejection of low-frequency baseline wander components; processed EEG signal decomposition using ACMD; rejection of mode containing OAs based on proposed peak point count threshold criteria. The performance of the proposed framework is assessed using noise-free EEG and EEG corrupted with OAs signals with different shapes and frequencies taken from three publicly available databases and in terms of different performance metrics. Further, comparative performance analysis demonstrates that the proposed framework outperforms the existing OAs removal techniques based on wavelet thresholding, variational mode decomposition (VMD), and Savitzky-Golay filter (SG-filter).