For an animated human face model to appear natural it should produce eye movements consistent with human ocular behavior. During face-to-face conversational interactions, eyes exhibit conversational turn-taking and agent thought processes through gaze direction, saccades, and scan patterns. We have implemented an eye movement model based on empirical models of saccades and statistical models of eye-tracking data. Face animations using stationary eyes, eyes with random saccades only, and eyes with statistically derived saccades are compared, to evaluate whether they appear natural and effective while communicating.

Smooth pursuit eye movements are guided largely by retinal-image motion. To compensate for neural conduction delays, the brain employs a predictive mechanism to generate anticipatory pursuit that precedes target motion (E. Kowler, 1990). A critical question for interpreting neural signals recorded during pursuit concerns how this mechanism is interfaced with sensorimotor processing. It has been shown that the predictor is not simply turned-off during randomization because anticipatory eye velocity remains when target velocity is randomized (E. Kowler & S. McKee, 1987; G. W. Kao & M. J. Morrow, 1994). This study was completed to compare pursuit behavior during randomized motion-onset timing with that occurring during direction or speed randomization. We found that anticipatory eye velocity persisted despite motion-onset randomization, and that anticipation onset time was between that observed in the different constant-timing conditions. This centering strategy was similar to the bias of eye velocity magnitude away from extremes observed when direction or speed was randomized. Such a strategy is comparable to least-squares error minimization, and could be used to facilitate acquisition of a target when it begins to move. Centering was in some observers accounted for by a shift of eye velocity toward that generated in the preceding trial. The results make unlikely a model in which the predictor is disengaged by randomizing stimulus timing, and suggest that predictive signals always interact with those used in sensorimotor processing during smooth pursuit.

Abstract

For an animated human face model to appear natural it should produce eye movements consistent with human ocular behavior. During face-to-face conversational interactions, eyes exhibit conversational turn-taking and agent thought processes through gaze direction, saccades, and scan patterns. We have implemented an eye movement model based on empirical models of saccades and statistical models of eye-tracking data. Face animations using stationary eyes, eyes with random saccades only, and eyes with statistically derived saccades are compared, to evaluate whether they appear natural and effective while communicating.