Document Type



This item is available under a Creative Commons License for non-commercial use only


Computer Sciences


Understanding the cognitive and functional behaviour of the brain by its electrical activity is an important area of research. Electroencephalography (EEG) is a method that measures and record electrical activities of the brain from the scalp. It has been used for pathology analysis, emotion recognition, clinical and cognitive research, diagnosing various neurological and psychiatric disorders and for other applications. Since the EEG signals are sensitive to activities other than the brain ones, such as eye blinking, eye movement, head movement, etc., it is not possible to record EEG signals without any noise. Thus, it is very important to use an efficient noise reduction technique to get more accurate recordings. Numerous traditional techniques such as Principal Component Analysis (PCA), Independent Component Analysis (ICA), wavelet transformations and machine learning techniques were proposed for reducing the noise in EEG signals. The aim of this paper is to investigate the effectiveness of stacked autoencoders built upon Gated Recurrent Unit (GRU) based Recurrent Neural Network (RNN) layers (GRU-AE) against PCA. To achieve this, Harrell-Davis decile values for the reconstructed signals’ signal-to- noise ratio distributions were compared and it was found that the GRU-AE outperformed PCA for noise reduction of EEG signals.