Real-Time Stress Level Detection System Using EEG Signal Processing and Deep Learning Techniques

Abstract

In recent years, electroencephalography (EEG) data for stress level detection has garnered significant attention due to its non-invasive nature and potential applications in various fields, including healthcare, psychology, and human-computer interaction. This paper presents a comprehensive framework for stress level detection utilizing electroencephalography (EEG) data and deep learning techniques, focusing on real-time prediction. Leveraging the Neuphony Flex Cap - 8 Channel Wireless EEG Head Cap System as the primary data source, the research aims to provide an integrated solution for stress level assessment, spanning data preprocessing, model training, evaluation, and deployment within a user-friendly graphical interface. The initial phase involves data preprocessing, where raw EEG signals transform, such as feature extraction and normalization, facilitating optimal data analysis. Model development utilizes convolutional neural networks (CNNs) and EEGNet for stress level classification, with strategies for addressing class imbalance and enhancing model robustness. Rigorous evaluation assesses model performance using various metrics, while visualization techniques offer insights into model dynamics. Deployment within a GUI application enables real-time stress level prediction, empowering users with seamless interaction and visualization of prediction outcomes. The presented framework offers scalability and data integrity through MySQL databases, with potential applications in healthcare and personalized stress management. Stress analysis is conducted across various EEG channels, each corresponding to different brain regions, to capture localized neural activity associated with stress.