Abstract:Mental stress poses a significant public health concern due to its detrimental effects on physical and mental well-being, necessitating the development of continuous stress monitoring tools for wearable devices. Blood volume pulse (BVP) sensors, readily available in many smartwatches, offer a convenient and cost-effective solution for stress monitoring. This study proposes a deep learning approach, a Transpose-Enhanced Autoencoder Network (TEANet), for stress detection using BVP signals. The proposed TEANet model was trained and validated utilizing a self-collected RUET SPML dataset, comprising 19 healthy subjects, and the publicly available wearable stress and affect detection (WESAD) dataset, comprising 15 healthy subjects. It achieves the highest accuracy of 92.51% and 96.94%, F1 scores of 95.03% and 95.95%, and kappa of 0.7915 and 0.9350 for RUET SPML, and WESAD datasets respectively. The proposed TEANet effectively detects mental stress through BVP signals with high accuracy, making it a promising tool for continuous stress monitoring. Furthermore, the proposed model effectively addresses class imbalances and demonstrates high accuracy, underscoring its potential for reliable real-time stress monitoring using wearable devices.