BCIwiki (bciwiki.com) — A non-invasive brain-computer interface that chains eye tracking and motor imagery into a two-step confirmation sequence reached up to 100% accuracy in tests with 15 healthy participants on a 16-channel EEG setup, outperforming motor imagery alone across every channel configuration. The work by Gowtham Reddy N, KongFatt Wong-Lin and Yogesh Kumar Meena was posted to the arXiv preprint server on August 22, 2026, and has not been peer reviewed. It was selected for the Brain-Machine Interface Systems Session at IEEE SMC 2026.
Motor imagery BCIs have long suffered from low discriminability and high inter-subject variability: users imagine a movement, the system reads the intent from EEG, but the signal shifts between people and between sessions, capping the number of commands that can be reliably separated. The researchers first examined how visual fixation affects the stability of neural responses, then proposed an asynchronous hybrid paradigm in which eye tracking makes the direct selection and a single motor imagery act confirms it, cutting the operational steps of conventional systems.
Motor imagery-related information was found to be localised mainly in motor cortex regions. A reduced-channel configuration using a support vector machine reached 0.58 accuracy, comparable to 0.54 for the full montage. The hybrid paradigm beat conventional motor imagery in all channel configurations, reaching up to 100% accuracy with greater robustness. The authors conclude that visual fixation improves neural response stability and that combining eye tracking with motor imagery could support reliable, scalable multi-command BCI systems for real-world use.