Stress Biomarkers for Applied Tasks: Analysis of Voice and Cardiac Signals

This paper reviews recent advances in the use of voice and cardiac signals as biomarkers for detecting psychological stress in applied settings. It outlines how vocal cues such as pitch, intensity, and articulation patterns change under stress, and how machine learning models can leverage these acoustic features for stress recognition. Similarly, it explores the utility of cardiac signals—particularly heart rate variability (HRV)—as indicators of autonomic nervous system activation during stress. The paper summarizes state-of-the-art methods for feature extraction, classification, and multimodal fusion, highlighting both classical approaches and emerging deep learning techniques. Challenges such as data scarcity, label reliability, and generalization to real-world scenarios are discussed. The findings support the development of adaptive, user-centered systems for early stress identification, with potential applications in safety-critical environments and wearable technologies.