Non-Intrusive Stress Detection Using Laptop Interactions

In this article, I share how I built a system to detect cognitive stress passively — without any wearables, sensors, or interrupting the user — simply by analyzing how someone interacts with their laptop.

The model leverages keystroke dynamics, mouse movement, and gaze tracking via the built-in webcam to identify signs of stress. One of the breakthroughs was the implementation of gaze detection using head pose estimation and facial landmarks, which led to a 30% improvement in accuracy, especially for users wearing glasses.

The system achieved up to 89% prediction accuracy and was awarded Best Project - Taught Master’s 🏆.

Read the article on Medium