Stress Detection Systems: A Comprehensive Review of Text, Speech, and Image-Based Approaches Using Machine Learning

The stress detection problem has attracted considerable attention among researchers in the field of modern medicine as stress may have a significant impact on the health condition of a person. However, the methods traditionally applied in detecting stress are based on physiological sensors and, thus, require special devices, high costs and invasive methods of collecting data that make them inappropriate for real-time use. To solve the stated problem, the paper suggests a multimodal stress detection system that utilizes text, speech and facial features as the input. In particular, this work uses transformer language models for processing the text, deep convolutional neural networks for identifying emotions from speech signals in the form of spectrograms using transfer learning approach and MobileNetV2 model for facial expression recognition. The performance of these methods is tested based on widely used benchmarks such as FER-2013 for facial expression analysis and RAVDESS for the speech emotion recognition.

© 2026 — Authors retain the copyright of this article. This is an open access article distributed under the Creative Commons Attribution License (CC BY 4.0) , which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.