🤖 AI Summary
This work addresses the functional mapping problem from electroencephalography (EEG) signals to eye-movement trajectories. We introduce EEG-EyeTrack, the first publicly available benchmark tailored for functional data analysis (FDA), incorporating data acquired from both consumer-grade and research-grade hardware. We formalize an FDA-specific task definition and evaluation paradigm, and propose a B-spline-based functional representation. Our framework integrates functional principal component analysis (FPCA), time-warping-aware regression, and functional neural networks (FNNs). Compared to conventional time-series models, our FDA-based baselines reduce the mean angular error in gaze trajectory reconstruction to 8.3° on consumer-grade EEG devices—marking a substantial improvement in decoding accuracy. This work fills a critical gap by establishing the first FDA-oriented benchmark for neurobehavioral decoding, and provides a reproducible, scalable methodological framework for EEG-driven functional gaze modeling.
📝 Abstract
A new benchmark dataset for functional data analysis (FDA) is presented, focusing on the reconstruction of eye movements from EEG data. The contribution is twofold: first, open challenges and evaluation metrics tailored to FDA applications are proposed. Second, functional neural networks are used to establish baseline results for the primary regression task of reconstructing eye movements from EEG signals. Baseline results are reported for the new dataset, based on consumer-grade hardware, and the EEGEyeNet dataset, based on research-grade hardware.