Unsupervised stress detection from remote physiological signal

Emotion detection from physiological signals is a relatively new classification task. This paper proposes a comparative study of unsupervised approaches to detecting stress from remote physiological signals. The aim of this study is to explore three techniques for unsupervised classification to assess and quantify mental stress states using a low-cost webcam. An interactive version of the Stroop color word test was employed to induce stress. The unsupervised classification techniques considered are k-means, the mixture of Gaussians model and Self organized map. The obtained results show that the best rate of stress detection is achieved using k-means clustering.