A novel few-channel strategy for removing muscle artifacts from multichannel EEG data

Abstract

Electroencephalography (EEG) recordings are often contaminated by muscle artifacts. Various methods have been proposed to suppress muscle artifacts from multichannel EEG recordings. However, the existing multichannel approaches have their own limitations. Instead of using multichannel techniques, in this paper, we propose an effective few-channel technique that combines multivariate empirical mode decomposition (MEMD) with canonical correlation analysis (CCA), termed as MEMD-CCA, to remove muscle artifacts from multichannel EEG recordings. The proposed method consists of two steps. First, the proposed method partitions multichannel EEG into several few-channel EEG groups and deals with each group individually. Next, MEMD is utilized to decompose every few-channel EEG groups into intrinsic mode functions (IMFs) and then CCA is applied on the IMFs to separate sources related to muscle activity. We compare the denoising performance between multichannel and few-channel approaches through simulated and real-life EEG data contaminated by muscle artifacts. The results demonstrate the advantage of few-channel approaches over multichannel ones for rejecting muscle artifacts without altering the desired EEG information.