Detecting Negative Emotional Stress Based on Facial Expression in Real Time

Abstract

Negative emotional stress can be seen as a physiological response to mental and physical challenges. Exposure to stressful situations with a long time can have adverse effects on people, such as depression, which finally results in suicide in severe case, so it is important to monitor stress in real time and treat it properly. In this paper, we propose a new framework for stress detection in real time. The framework detects stress by recognizing three stress related facial expressions, anger, fear and sadness. We also propose a connected convolutional network, which combines low-level features with high-level features to train the deep network to recognize facial expressions. If the number of stress related frames exceeds a threshold value, the framework will remind people to take a break to relax. The experiment results demonstrate that our proposed method has better performance on facial expression recognition and realizes high-performance stress detection.