Personalized cognitive state detection for smart eyewear

Stress and cognitive load affect performance and wellbeing in many everyday situations, from driving to demanding work environments. Detecting these states early could allow devices to step in and help before problems arise.

This dissertation shows how sensors that could be built into smart eyewear, such as eye tracking cameras alongside heart and skin sensors, can detect stress and mental workload. The research shows that combining eye measurements with cardiac and electrodermal signals reliably distinguishes between different stress and workload states, with eye based signals proving especially valuable.

A key challenge in state detection is that everyone's body responds to the demands differently, so detection models often need extensive personal data to work well. This dissertation presents new methods that personalize the model using only a few minutes of baseline measurement, cutting the data requirement while still improving accuracy over generic models built for everyone. The study found comparable performance between a lightweight feature-based approach with data-constrained personalization, and a modern artificial intelligence foundation model.

The findings offer smart eyewear and wearable technology developers a practical guideline for building applications that adapt to a person's stress or workload in real time, with potential uses spanning safety critical work, driving, learning, and personalized digital health tools.