TheunscentedKalmanfilterisadevelopedwell-knownmethodfornonlinearmotionestimationandtracking.However,thestandardunscentedKalmanfilterhastheinherentdrawbacks,suchasnumericalinstabilityandmuchmoretimespentoncalculationinpracticalapplications.Inthispaper,wepresentanovelsamplingstrongtrackingnonlinearunscentedKalmanfilter,aimingtoovercomethedifficultyinnonlineareyetracking.Intheaboveproposedfilter,thesimplifiedunscentedtransformsamplingstrategywithn+2sigmapointsleadstothecomputationalefficiency,andsuboptimalfadingfactorofstrongtrackingfilteringisintroducedtoimproverobustnessandaccuracyofeyetracking.ComparedwiththerelatedunscentedKalmanfilterforeyetracking,theproposedfilterhaspotentialadvantagesinrobustness,convergencespeed,andtrackingaccuracy.Thefinalexperimentalresultsshowthevalidityofourmethodforeyetrackingunderrealisticconditions.
