Real-Time Mental Stress Monitoring Using Machine Learning On User Interaction Patterns
Year: 2026
Type: article
Abstract: The detection of mental stress is becoming an essential function of modern Human Computer Interaction (HCI) systems due to its effects on user performance and mental well-being. This work introduces a novel real-time system to monitor mental stress by using machine learning techniques on non-invasive behavioral biometrics. The system collects data from user interaction through keyboard dynamics (typing speed and latency), mouse behavior (jitter and click frequency) and eye-tracking (using Eye Aspect Ratio, EAR, derived by using Media Pipe). We employ a Long Short-Term Memory (LSTM) model in order to capture the temporal patterns within the user's behaviour and the Isolation Forest algorithm for anomaly detection by identifying the discrepancies with the user's individual baseline pattern. A user-specific modelling approach with user-specific normalization is integrated into the system to adapt the model to the specific characteristics of each user. The proposed system employs an efficient SQLite lightweight database to store sessions and update the model accordingly. Our results show better performance based on F1-score and more resilience towards biased behavioral data than other approaches while keeping the low latency characteristics of the proposed system which make it a good candidate for real-time monitoring applications.
Language: English
Topic: Emotion and Mood Recognition
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7162201296