University of Tabriz Tabriz Journal of Electrical Engineering 2008-7799 2026 04 11 Evaluation of Nonlinear Characteristics in Multi-Wavelength Photoplethysmography Signals for Muscle and Skin Sympathetic Nerve Activity in Stress Classification Evaluation of Nonlinear Characteristics in Multi-Wavelength Photoplethysmography Signals for Muscle and Skin Sympathetic Nerve Activity in Stress Classification 21393 10.22034/tjee.2026.68024.5044 FA Reyhaneh Boskabadi Bachelor’s student, Department of Biomedical Engineering, Faculty of Engineering, Imam Reza International University, Mashhad, Iran Asmae Zoghi Bilondi Bachelor’s student, Department of Biomedical Engineering, Faculty of Engineering, Imam Reza International University, Mashhad, Iran Sara Moshiryan Master&#039;s student, Department of Biomedical Engineering, Faculty of Engineering, Imam Reza International University, Mashhad, Iran Arezoo Sanati Fahandari Master&#039;s student, Department of Biomedical Engineering, Faculty of Engineering, Imam Reza International University, Mashhad, Iran Ateke Goshvarpour Department of Biomedical Engineering, Faculty of Engineering, Imam Reza International University, Mashhad, Iran Journal Article 2025 07 09 Stress is one of the common psychological challenges of the 21st century that affects the function of the autonomic nervous system. In this study, the feasibility of non-invasive monitoring of muscle sympathetic nerve activity (MSNA) and skin sympathetic nerve activity (SSNA) was investigated using pre-extracted nonlinear features from multi-wavelength photoplethysmography (PPG) signals and by employing machine learning algorithms. PPG data from 32 healthy individuals (19 to 38 years old) at four wavelengths (red, infrared, blue, and green) were analyzed during three time phases (pre-stress, during stress, and post-stress induced by handgrip and cold pressor tests). Nonlinear features, including Higuchi, Katz, and Petrosian fractal dimensions, approximate entropy, and sample entropy, were used for classification. These features were applied in three classifiers—Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Recurrent Neural Network (RNN)—to classify stress states. The highest performance was observed using approximate entropy at the blue wavelength when distinguishing the pre-stress phase from the first two minutes of stress in the handgrip test (SVM: accuracy = 90.96%, AUC = 1.00). Independent t-tests and Wilcoxon tests revealed significant differences (p < 0.05) in the blue, green, and infrared wavelengths. These results confirm the role of nonlinear features and optimal wavelength selection in effective stress monitoring and highlight the potential application of PPG as a low-cost, non-invasive tool. Stress is one of the common psychological challenges of the 21st century that affects the function of the autonomic nervous system. In this study, the feasibility of non-invasive monitoring of muscle sympathetic nerve activity (MSNA) and skin sympathetic nerve activity (SSNA) was investigated using pre-extracted nonlinear features from multi-wavelength photoplethysmography (PPG) signals and by employing machine learning algorithms. PPG data from 32 healthy individuals (19 to 38 years old) at four wavelengths (red, infrared, blue, and green) were analyzed during three time phases (pre-stress, during stress, and post-stress induced by handgrip and cold pressor tests). Nonlinear features, including Higuchi, Katz, and Petrosian fractal dimensions, approximate entropy, and sample entropy, were used for classification. These features were applied in three classifiers—Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Recurrent Neural Network (RNN)—to classify stress states. The highest performance was observed using approximate entropy at the blue wavelength when distinguishing the pre-stress phase from the first two minutes of stress in the handgrip test (SVM: accuracy = 90.96%, AUC = 1.00). Independent t-tests and Wilcoxon tests revealed significant differences (p < 0.05) in the blue, green, and infrared wavelengths. These results confirm the role of nonlinear features and optimal wavelength selection in effective stress monitoring and highlight the potential application of PPG as a low-cost, non-invasive tool.