Single-trial Classification of Fixation-related Potentials in Guided Visual Search Tasks using A Riemannian Network

Abstract

Brain responses to visual stimulus can provide information about cognitive process or intentions. Several studies show that it is feasible to use stimulus-dependent modulation of the evoked brain responses after gaze movements (i.e., Fixation Related Potential, FRP) to predict the interested object of human. However, the performance of the state-of the-art shallow models for FRP classification is still far from satisfactory. Recent years, Riemannian geometry based on deep learning has gained its popularity in many image and video processing tasks, thanks to their ability to learn appropriate statistical representations while respecting Riemannian geometry of the data in such fields. In this paper, we have investigated a Riemannian network for classifying FRP in guided visual search task. Experiment results showed that the Riemannian network improved classification performance significantly in comparison to the shallow methods.