Remove Motion Artifacts from Scalp Single Channel EEG based on Noise Assisted Least Square Multivariate Empirical Mode Decomposition

Abstract

Noninvasive scalp single channel EEG is increasing being applied in our daily lives, due to its minimal instrumentation complexity and safety compared with multichannel EEG and invasive EEG. The unavoidable artifacts really hamper its applications and the artifacts correction remains challenging especially in the case of only one channel recordings available. In this paper, we propose a novel approach for removing motion artifacts, particularly frequent during recording, from scalp single channel EEG recordings. The novel approach is developed based on Noise Assisted Least Square Multivariate Empirical Mode Decomposition (NALSMEMD), which solves the problems of subspace incompleteness in Ensemble EMD (EEMD) and therefore further improve the motion artifacts removal performance. First, the single channel EEG is decomposed into several Intrinsic Mode Functions (IMFs) assisted by the separated white Gaussian noise channels. Then the artifacts related IMFs are selected and rejected according to the IMFs' autocorrelation coefficients. Finally, the EEG related IMFs are reconstructed as the motion artifacts free EEG. The 23 sessions of single channel EEG data downloaded from https://www.physionet.org/content/motion-artifacts/1.0.0/ are used in our study for verifying the performance of our approach. The results show that our approach outperforms EEMD based approach in terms of SNR change before and after artifacts removal and percentage reduction in artifacts after artifacts removal.