Efficient automated method to extract EOG artifact by combining Circular SSA with wavelet and unsupervised clustering from single channel EEG

Abstract

The emergence of affordable single-channel electroencephalography (EEG) headbands has introduced new possibilities for applications in brain–computer interface (BCI) systems and health monitoring. However, these EEG signals are often contaminated by electrooculogram (EOG) activity caused by eyelid blinking. This poses a challenge, especially in portable EEG devices that typically have limited or single EEG channels. The presence of eye-blink artifacts can mislead the diagnosis of brain state activity. Consequently, there is a growing demand to extract eye-blink artifacts from single-channel EEG signals. However, during the process of removing these artifacts from contaminated EEG signals, low-frequency artifact components in non-artifact regions are also inadvertently eliminated. In this work, an automated, robust and efficient framework is presented by combining Circular Singular Spectral Analysis (Ci_SSA) with wavelet analysis and unsupervised clustering. This framework aims to remove artifact segments while preserving the low-frequency content in non-artifact regions. Simulations were carried out on both synthetic and real data, and the results were compared with alternative methods. The performance of the proposed method is evaluated by metrics such as Relative Root Mean Square Error (RRMSE), Mean Absolute Error (MAE), Correlation Coefficient (CC) and Artifact Rejection Ratio (ARR) for synthetic signals, as well as Signal-to-Error Ratio (SER) and Artifact-to-Residue Ratio (AReR) for real EEG signals. After applying the proposed algorithm to filter the contaminated EEG signal, a noticeable enhancement in the detection accuracy of driver fatigue is observed as compared to the pre-filtering stage.