Noise of Eye Movement Detection by Videooculography

Abstract

The noise of pupil center detection algorithms used in videoculography was investigated. Proposed a novel performance parameter, which characterize the noise of algorithms. This parameter characterize true algorithm noise better than root mean squared (rms) value or low pass filter. Investigated pupil contour filtering algorithms: line and circle filtering. Coordinates averaging algorithm has a highest noise level. Combination of this algorithm and pupil contour filtering algorithms reduces noise level. Circle approximation algorithm fails to locate pupil center in high noise conditions. The pixel resolution (8-bit or 10-bit) doesn’t have a influence to noise level. Ill. 10, bibl. 3 (in English; summaries in Lithuanian, English, Russian).