Application of a Novel Neural Approach to 3D Gaze Tracking: Vergence Eye-Movements in Autostereograms

Abstract

Vergence eye-movements occur not only in real environ-\nments, but also in virtual 3D environments. Autostere-\nograms can cover large visual angles without requiring\nvergence beyond natural parameters and are thus well-\nsuited for the investigation of vergence movements in\nvirtual 3D images. We developed an anaglyph-based\n3D calibration procedure and used a\nparametrized self-\norganizing map\n(PSOM) to approximate the 3D gaze\npoint from a subject's binocular eye-movement data.\nBesides analyzing the general pattern of vergence eye-\nmovements in autostereogram images, the present study\nexamined the in∞uence of image granularity on these\nmovements. Unlike previous research on random-dot\nstereograms, we found substantially overshooting con-\nvergence eye-movements, especially for medium gran-\nularities. Moreover, divergence movements were com-\npleted more quickly for coarse than for Øne granularities.\nResults are discussed in the context of granularity eÆects\non autostereogram perception and the dissociation be-\ntween convergence and divergence eye-movements.