Neural Spectre: AI/ML-driven imperceptible brain computer interfaces for extended reality

Source: https://engineering.cmu.edu/education/undergraduate-studies/undergraduate-research/honors-research/2025/chan-neural-spectre.html Parent: https://engineering.cmu.edu/education/undergraduate-studies/undergraduate-research/honors-research/ece.html

Imagine navigating a sophisticated user interface on smart glasses while on the go using only your eyes and attention, without the use of tiring or awkward hand gestures and voice commands.\ \ SSVEPs (steady-state visual evoked potentials) are a robust brain-computer interaction that can enable screen interaction using only visual attention and minimal user training. They have been successfully used in high-precision applications including virtual keyboards and robotic control with 97-99% accuracy.\ \ When a user focuses on a flickering target, the visual cortex automatically generates detectable brain waves (EEG) at the same frequency, allowing real-time, hands-free selection.\ \ However, conventional SSVEP systems use flicker frequencies that can cause substantial visual fatigue.\ \ Here, we propose an imperceptible SSVEP-based BCI tailored for extended reality (XR) devices like smart glasses and headsets that leverage AI-generated flicker patterns that are both imperceptible to the human eye and which maximizes information transfer allowing high bandwidth communication between brains and machines.