Artifact Suppression from EEG Signal Using Sub-band Thresholding Approach

EEG recordings are typically influenced by different artifacts originating from non-neural sources, complicating subsequent precise signal classification. The reliable detection and removal of artifacts from EEG signals using an automated signal processing technique is a prominent study domain. This study presents a wavelet-based approach for the suppression of artifacts in EEG data, which identifies ideal threshold values to enhance artifact removal efficacy. In the suggested algorithm, iterated over the threshold settings until optimal accuracy or minimal distortion is attained, utilizing a reference dataset for decision making. The criteria for optimum selection rely on matrices that measure the signal-to-noise ratio, mean square error, and other factors. The technique is evaluated on a genuine dataset of EEG signals containing ocular artifacts. The results indicate a 16.93 dB enhancement in the signal-to-noise ratio (SNR), confirming that adaptively determining optimal threshold parameter values yields superior performance compared to using any predefined threshold parameters. This research will provide the EEG signal analysis community with a platform to further investigate the issue of selecting wavelet settings effectively