Stochastic models of gaze-shifts

Stochastic models of gaze-shifts

One point that is not addressed by the great majority of computational models is the "noisy", idiosyncratic variation of the random exploration exhibited by different observers when viewing the same scene, or even by the

same subject along different trials.

Such variations speak of stochastic nature of scanpaths - the succession of gaze-shifts - , which is particularly evident for those resulting from saccadic eye movements

The problem: how random are gaze-shifts?

Eye-tracked scanpaths of different subjects

looking at the same emotional expression

Eye-tracked scanpaths of different subjects

looking at the same video (CNRS dataset)

This research work was initially inspired by seminal work of Prof. Larry Stark, Telerobotics and Neurology Units, University of California Berkeley.

This research has been carried on with Prof. Mario Ferraro (University of Torino)

A research (non random) walk

Random walk

Brownian Random walk

Levy Random walk

The Foraging

metaphor

Constrained

Levy search

Levy

Hybrid

Montecarlo

Composite

walks

Active

random

sampling

Alpha-stable

distributions

Motor behavior

Simple

composite

search

Physica A

(2004)

Euvip

(2010)

ICIAP

(2011)

Gaze-shift mechanisms can be conceived as a motor program implementation of an active random sampling strategy that the Human Visual System has evolved in order to efficiently and effectively infer properties of the surrounding world

Gaze-shifts are generated by an underlying stochastic process.

If observed gaze-shifts are generated by an underlying stochastic process the distribution functions and the temporal dynamics of eye movements should be completely specified by the stochastic process

Working assumptions

Euvip

(2011)