Stress Detection Based On Facial Expression

Abstract

 Stress has become a major concern in today’s fast-paced lifestyle, affecting mental health and overall well-being. Traditional stress detection methods rely on self-reporting techniques or physiological sensors, which are intrusive, expensive, and not suitable for continuous real-time monitoring. This paper presents a real-time facial emotion and stress detection system using computer vision and deep learning techniques. The system captures facial images through a webcam and detects faces using the Haar Cascade classifier. The detected face is preprocessed through grayscale conversion, resizing, and normalization to enhance model performance. A Convolutional Neural Network (CNN) is used to extract important facial features, and a Deep Neural Network (DNN) with a Soft max classifier predicts the emotion along with confidence scores. Based on the predicted emotion and confidence level, a rule-based algorithm categorizes stress levels into Low, Medium, or High. The system also generates suitable recommendations to help users manage stress effectively. The proposed model provides a non-intrusive, affordable, and real-time solution for early stress detection and mental health monitoring applications.