Intelligent approach for artifacts removal from EEG signal using heuristic-based convolutional neural network

Abstract

In general, the electrical activity of the brain is recorded using Electroencephalography (EEG), which is contaminated with some signal artifacts. By using automatic removal of artifacts from EEG signals, different Brain-Computer Interface (BCI) and clinical diagnostics applications are in practice. However, they are not efficient to remove the artifact from the EEG signal. Thus, we plan for the intelligent model for artifacts removal of EEG signal. The two main phases of the proposed model are training and testing. The deep learning model in the training phase is used as the filter to automatically remove noise from the contaminated EEG signal. The proposed model adopts improved One-Dimensional Convolution Neural Networks (1D-CNN) for artifacts removal from EEG signals. Here, a new hybrid algorithm named Spider Monkey-based Electric Fish Optimization (SM-EFO) is proposed by integrating the Spider Monkey Optimization (SMO) and Electric Fish Optimization (EFO) algorithm. The model parameters of the One-Dimensional Convolutional Neural Networks (1D-CNN) are tuned by using SM-EFO. The experimentation is performed on the standard benchmark dataset, and the experimental results establish that the proposed model can achieve significant improvement and get cleaner waveforms in terms of several performance measures when compared to the conventional models.