Optokinetic nystagmus (OKN) is the sawtooth movement of the eye elicited when an observer views a repeated moving pattern. We present a method for identifying the presence and direction of OKN in recordings of the eye made using a standard off-the-shelf video-camera or webcam. Our approach uses vertical edge detection to determine the limbus/iris boundary, and we estimate the velocity of the edge using Lucas-Kanade optical flow. Heuristic rules are applied to identify saccadic velocity peaks from the resulting velocity signal. The normalized average of the resulting peaks is used to estimate the presence/direction of OKN. Our preliminary testing with six participants observing global motion stimuli with full or partial coherence yields an accuracy of 93% which compares favorably to the performance of an experienced human observer (98% accuracy). Additional tests using high contrast, square-wave gratings show that performance of the technique is consistent at stimulus speeds of 5 and 10deg/s and that OKN is not reported by the algorithm when participants view stationary stimuli.
Keywords: Optic flow; Optokinetic nystagmus; Random-dot-kinetogram; Video-oculography.
Copyright © 2014 Elsevier Ltd. All rights reserved.