A novel end-to-end 1D-ResCNN model to remove artifact from EEG signals

Abstract

Electroencephalography (EEG) signals are an important tool in the field of clinical medicine, brain research and the study of neurological diseases. EEG is very susceptible to a variety of physiological signals, which brings great difficulties to the research and analysis of EEG signals. Therefore, removing noise from EEG signals is a key prerequisite for analyzing EEG signals. In this paper, a one-dimensional residual Convolutional Neural Networks (1D-ResCNN) model for raw waveform-based EEG denoising is proposed to solve the above problem. An end-to-end (i.e. waveform in and waveform out) manner is used to map a noisy EEG signal to a clean EEG signal. In the training stage, an objective function is often adopted to optimize the model parameters and in the test stage, the trained 1D-ResCNN model is used as a filter to automatically remove noise from the contaminated EEG signal. The proposed model is evaluated on the EEG signal from the CHB-MIT Scalp EEG Database, and the added noise signals are obtained from the database. We compared the proposed model with the independent of the composite analysis (ICA), the fast independent composite analysis (FICA),Recursive least squares(RLS) filter,Wavelet Transform (WT) and Deep neural network(DNN) models. Experimental Results show that the proposed model can yield cleaner waveforms and achieve significant improvement in SNR and RMSE.Meanwhile, the proposed model can also preserve the nonlinear characteristics of EEG signals.