A Revisit Of Different Methods For Estimating Saccadic Velocities And Accelerations

Abstract

Purpose: : To evaluate a suggested method for evaluating the peak velocities and accelerations of saccades.

Methods: : We recorded 8 normal subjects at 240 Hz using infra–red videooculography with an accuracy to 0.1 degrees. Each subject performed over 1300 horizontal saccades at amplitudes varying from 5 to 20 degrees. Saccades with blinks within 100 msec of the saccade as well as those glissadic components greater than 25% of the total saccade were eliminated. This left more than 150 saccades for each subject. Peak velocity, acceleration and deceleration were evaluated using different methods. These include the standard method of a backwards difference, central difference (the most commonly used estimating algorithm), and an 8^th order approximation utilizing FIR filters. The main sequences generated were compared for each subject.

Results: : All calculations using backwards difference was decreased compared to other methods. Central difference showed a consistent decrease in velocity, averaging 15deg/sec less than that calculated by our approximation. More significant was the difference found in the absolute accelerations. The differences averaged 10000 deg/sec/sec. These were consistent amongst all subjects. Deceleration differences were similar.

Conclusions: : The method of estimating saccadic velocities have a significant effect on peak velocity but a much greater effect on measurements of acceleration and deceleration . As people analyze acceleration and deceleration as an important part of the analysis the method of estimation becomes critical. We recommend the use of higher order approximating algorithms such as the 8^th order approximation used here. We did not evaluate other approximations due to limitations of our mathematical software.

Keywords: eye movements • eye movements: saccades and pursuits