High-speed ocular artifacts removal of multichannel EEG based on improved moment matching

Abstract

Objective. The excellent signal-to-noise ratio (SNR) is the premise of electroencephalogram (EEG) research and applications. This study aims to use innovative method to swiftly remove the ocular artifacts (OAs) from multichannel EEG to enhance the SNR. Approach. The moment matching method which is prevalently used to removing stripe noise from hyperspectral images is adapted and improved to deduct OAs from EEG. This modified approach regards sampling points of multichannel EEG as pixels in images. It utilizes the propagation characteristics of EEG to correct the potential shift caused by OAs. Main results. By using mathematical derivation and empirical corroboration, the results suggest that the improved moment matching (IMM) is capable of reducing OAs effectively and reserving the EEG waveform information on the greatest extent compared to existing methods, such as independent component analysis (ICA) and second-order blind identification. In the frontal region heavily affected by OAs, the SNR increased by 138.1% compared with ICA, the whole SNR increased by an average of 58.7%. Moreover, low latency superiority is provided for real-time and offline processing. IMM is effective for OAs removal and it is helpful to improve SNR of multichannel EEG. Significance. IMM affords option offering preponderance for enhancement of SNR of multichannel EEG.