Prediction of viewed object sizes using features of visual evoked potentials and oculo-motors

In order to develop a type of automatic viewing assistance system, the prediction performance of viewing six sizes of Landolt circles was measured using pupil diameters, eye movements and the viewer's occipital single channel potentials. To reduce the influence of artefacts produced by these signals, empirical mode decomposition and other appropriate processing techniques were applied. Discriminant prediction of the visual sizes of stimuli was conducted using mathematical tools and feature sets extracted from these signals. The prediction performance increased when these feature sets were combined to optimise performance. Also, differences in the performance between stimulus sizes and individual subjects were determined.