A NOVEL NEURAL EEG DECODING FOR STRESS DETECTION USING OPTIMIZED GATED RECURRENT NEURAL NETWORKS

Abstract

Electroencephalography (EEG) is a non-invasive method used to record evoked potentials and electrical activity from the brain. Recently, Artificial Intelligence particularly machine learning (ML) and deep learning (DL) has gained brighter light of research for analyzing the different EEG patterns. Furthermore, decoding of stress from the EEG patterns using ML and DL algorithms is considered as the one of the brightest light of problem that remains challenge among the researchers. In this context, this paper proposes the Modified Gated Recurrent Neural networks (M-GRU) with the Optimized training neural network to decode the stress from EEG patterns with high accuracy and less computational complexity. The proposed GRU networks are used to extract the temporal features while the optimized learning models are used for the complexity free classification and detection which can be used for the further treatment and diagnosis process. The extensive experimentation is carried out using the PhysioBank EEG Stress datasets and various performance metrics such as accuracy, precision, recall, specificity and F1-score are calculated and compared with the other state-of -art learning models. Simulation results shows that the proposed algorithm has shown the better performance than the existing models and has gained the substantial attention in decoding EEG signals for stress detection.