Enhancing EEG-based Stress detection: Integrated techniques and optimization strategies

Abstract

This research presents an integrated approach to EEG-based stress detection, combining various signal processing techniques to offer a novel perspective on stress-related EEG signal analysis. The study explores spectral analysis, time-frequency feature extraction, Discrete Wavelet Transform (DWT), and optimization methods, including the use of Chirp Cosine Raised Window (CCRW) with Short-Time Fourier Transform (STFT). An advanced fusion model is introduced, integrating Bidirectional Long Short-Term Memory (BiLSTM) layers and a Transformer architecture to capture temporal patterns and global context awareness within EEG signals. The optimization strategies used for feature selection, enhance the model's efficiency and accuracy in real-world applications. Additionally, the effectiveness of employing CCRW with STFT for spectral analysis is demonstrated, leading to a more precise representation of EEG signals during stress-related activities. This research offers a road map for researchers and practitioners, emphasizing the synergistic fusion of diverse approaches to improve the accuracy and reliability of EEG-based stress detection.