Academic journal article Attention, Perception and Psychophysics

Fixation Identification: The Optimum Threshold for a Dispersion Algorithm

Academic journal article Attention, Perception and Psychophysics

Fixation Identification: The Optimum Threshold for a Dispersion Algorithm

Article excerpt

It is hypothesized that the number, position, size, and duration of fixations are functions of the metric used for dispersion in a dispersion-based fixation detection algorithm, as well as of the threshold value. The sensitivity of the I-DT algorithm for the various independent variables was determined through the analysis of gaze data from chess players during a memory recall experiment. A procedure was followed in which scan paths were generated at distinct intervals in a range of threshold values for each of five different metrics of dispersion. The percentage of points of regard (PORs) used, the number of fixations returned, the spatial dispersion of PORs within fixations, and the difference between the scan paths were used as indicators to determine an optimum threshold value. It was found that a fixation radius of 1° provides a threshold that will ensure replicable results in terms of the number and position of fixations while utilizing about 90% of the gaze data captured.

(ProQuest: ... denotes formulae omitted.)

1. Introduction

Cognitive processing takes place during fixations and allows the human to "see" (Just & Carpenter, 1984), but the eyes are never still. There is constant tremor of the eyes in the form of nystagmus, drifts, and microsaccades (Rayner, 1998). "A fixation . . . may be thought of as the mean X and Y position coordinates measured over a minimum period of time during which the eye does not move more than some maximum amount. More simply stated, point-of-gaze must continuously remain within a small area for some minimum time" (Eyenal, 2001, p. 12).

Irwin (1992) found the theoretical minimum duration for a single fixation to be 150 msec, whereas Manor and Gordon (2003) argued that 100 msec can also be justified. Rayner (1998) indicated that the mean duration of a single fixation may depend on the nature of the task. The mean fixation duration during silent reading is about 225 msec. During visual search, however, the mean fixation duration is 275 msec, and for tasks that require hand-eye coordination, such as typing, the mean fixation can be 400 msec (Rayner, 1998).

Saccades are rapid eye movements with velocities as high as 500?/sec used to reposition the central focus area of the eye (fovea) to a new location (Rayner, 1998). The duration of a saccade is influenced by the distance covered. For example, a 2? saccade, which is typical during reading, lasts about 30 msec, whereas a 5? saccade (typical of scene perception) lasts about 40-50 msec (Rayner, 1998).

The fixation, as the fundamental unit of eyetracking analysis, depends on both the algorithm used to separate fixations from saccades and the parameters employed for the algorithm (Shic, Chawarska, & Scassellati, 2008). Results produced by different algorithms may vary significantly (Spakov & Miniotas, 2007). One of the greatest restrictions of the available algorithms for fixation detection is the fact that the parameter settings are crucial. It has been shown by Shic et al. (2008), as well as Manor and Gordon (2003), that a change in the parameters may cause a change in the reported fixation durations. Choices made by researchers during analysis can lead to different interpretations of the same eyetracking data (Shic et al., 2008).

In this article, the dispersion threshold algorithm for fixation identification (I-DT) of Salvucci and Goldberg (2000) is implemented, and an attempt is made to determine the optimum dispersion threshold for each of four different measures of dispersion. Various characteristics of the identified fixation sequences, also known as scan paths, are analyzed at a range of thresholds. The optimum threshold for each measure is considered to be the one at which a number of indexes (the percentage of PORs included, the number of fixations identified, the position of the fixations, and their spatial distribution) are optimized to such an extent that the position and duration of a viewer's actual gaze is represented as accurately as possible. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.