Live Demonstration: Wearable Edge-AI Meets Real-Time Saccadic Eye Movement Classification

This live demonstration presents GAPSES, an ultra-low-power smart glasses platform for wearable, fully dry acquisition of electrooculography (EOG) and electroencephalography (EEG). Powered by a parallel RISC-V processor (GAP9), the system enables real-time, on-device edge-AI processing for robust and continuous human-machine interfaces. By performing edgeAI processing of biosignals directly onboard, GAPSES eliminates the need for streaming sensitive raw data off-device, significantly enhancing energy efficiency, improving user privacy, and removing the dependency on a stable wireless data connection. In this demo, GAPSES runs onboard a lightweight convolutional neural network (CNN) for fast classification of saccadic eye movements. The CNN outputs are then used for a cursor-control task, measuring the user's performance in bits per second. Classification outputs and raw data are also displayed in a dedicated GUI to increase user interest in the technology. A leaderboard tracks each participant's performance to enhance engagement, encouraging attendees to 'beat the score' by achieving fast and accurate control using their eyes. Our demo underscores the critical role of embedded AI in wearable biosignal processing while allowing participants to experience first-hand how bioaware interactions can be realized in a user-friendly, secure, and responsive wearable form factor.