OICC Press International Journal of Biophotonics and Biomedical Engineering (IJBBE) 2980-9037 6 1 2026 06 30 Real-Time Stress Detection Using Combined PPG and PCG Signals with Deep Learning 10.57647/ijbbe.2026.0601.03 EN Sanaz Dalvandi Department of Electrical Engineering, Isf.C., Islamic Azad University, Isfahan, Iran Atefeh Salimi Shahraki Department of Electrical Engineering, Laser and Biophotonics in Biotechnologies Research Center, Isf.C., Islamic Azad University, Isfahan, Iran https://orcid.org/0000-0001-7798-2928 Farhad Azimifar Department of Biomedical Engineering, Isf.C., Islamic Azad University, Isfahan, Iran Mahdi Sajadieh Department of Electrical Engineering, Isf.C., Islamic Azad University, Isfahan, Iran https://orcid.org/0000-0003-1445-1593 Journal Article 2026 06 30 Given the serious risks psychological stress poses to both mental and cardiovascular health, there is a need for real-time and unobtrusive monitoring methods. We present a new deep learning framework that uses synchronized photoplethysmogram (PPG) and phonocardiogram (PCG) inputs to classify stress.  In contrast to current methods that use multimodal sensors or electrocardiography (ECG), our system uses only PCG and PPG, which are inexpensive, noninvasive, and wearable technology compatible.  The architecture incorporates a Bi-LSTM for sequential dependency modeling and a 1D U-Net encoder for multiscale temporal feature extraction.  A physiologically interpretable, self-supervised learning strategy is made possible by the automatic derivation of stress labels from PPG-based inter-beat intervals using RMSSD (root mean square of successive differences).  State-of-the-art accuracy (98.5%) and F1-scores (>0.97) are demonstrated via experimental results on a public dataset, together with real-time inference capabilities (<50 ms on CPU). Our approach opens the door for scalable wearable health monitoring devices by proving the feasibility of interpretable, end-to-end deep learning for stress detection using synchronized acoustic and optical cardiac inputs.