Learning behavior of the eye fixation control system

Abstract

While studying the responses of the oculomotor system to horizontally moving visual stimuli some interesting learning patterns were observed. Responses to Gaussian random noise, sine, and square-wave inputs were investigated. Based on small-signal, linear analysis, the system's transfer function was computed on the basis of both sinusoidal and random inputs. Assuming unity feedback, the open-loop transfer function was also computed for both of the above inputs. It is shown that the anticipatory behaviors observed during the tracking of periodic inputs, can be attributed to a prediction operator. This operator is determined by comparing the open-loop transfer functions for both stimuli. It is shown that the effect of this operator is a significant increase in low-frequency gain, and considerable phase-lead; monotonically increasing with frequency.