Detection of Eye Movement Characteristics Using Reservoir Computing in High-Noise Environments

Accurate real-time detection of saccades is crucial in neuroscience but challenging under high noise levels. We introduce a model based on reservoir computing and tested on simulated eye movement signals with varying $1 / \mathrm{f}$ noise. The model achieved 99.5% accuracy without noise and maintained over $\mathbf{9 0 \%}$ accuracy even at high noise levels, demonstrating its noise resistance and suitability for real-time brain-computer interfaces.