Saccade Identification During Driving Simulation from Eye Tracker Data with Low-Sampling Frequency

This paper introduces an improved velocity-based method with optimal velocity threshold parameter for saccade detection to calculate saccade number per second, saccade duration, and velocity peak from the eye tracker data during driving simulation. The algorithm is tested on data recorded with low-sampling frequency of 50 Hz and on the interpolated with sampling frequency of 200 Hz. The obtained results for saccade-related features are in the expected ranges reported in literature. However, results showed statistically significant changes in all parameters (except in standard deviation of saccade duration) as a consequence of interpolation to 200 Hz. Moreover, slight change in threshold parameter (from 5.75 to 6.0) for interpolated data led to statistically significant changes in all velocity-related saccade features. Although results showed that selection of processing technique plays a major role in the analysis of eye tracker data and influences saccade parameters, a more thorough algorithm evaluation is required for guided threshold estimation for saccade identification from eye tracker data recorded with low sampling frequency.