Visual Search with selective tuning
Evgueni Simine, Antonio J. Rodríguez-Sánchez, John K. Tsotsos · Journal of Vision · 2010
Visual attention involves much more than simply the selection of next fixation for the eyes or for a camera system. Selective Tuning (ST) (Tsotsos et al. 1995;2005) presents a framework for modeling this broader view of attention and in this work we show how it performs in covert visual search tasks by comparing its performance to the same input visual displays that human subjects have seen and qualitatively comparing the model's performance to human performance. Two implementations of ST have been developed. The Motion Model (MM) recognizes and attends to motion patterns and the Object Recognition Model (ORM) recognizes and attends to simple objects formed by the conjunction of various features. Two experiments were carried out in the motion domain. A simple odd-man-out search for CCW rotating octagon among identical CW rotating octagons produced linear increase in search time with the increase of set size. The second experiment was similar to one described in Thornton and Gilden(2001) paper and produced qualitatively similar results. The validity of the ORM was first tested by successfully duplicating the results of Nagy and Sanchez(1990). Our second experiment aimed at an evaluation of the model's performance for the feature-conjunction-inefficient continuum search slopes (Wolfe, 1998). For conjunction search we followed Bichot and Schall(1999) (find a red circle among green circles and red crosses). For feature search ORM looked for a circle among crosses and for inefficient search we simulated Egeth and Dagenbach(1991). Inefficient search produced a slope of 0.49, followed by conjunction search (slope of 0.36) and feature search was practically flat (slope of 0.00), these results show the same kind continuum of search slopes as described by Wolfe(1998). We conclude that ST provides a valid explanatory mechanism for human covert visual search performance, an explanation going far beyond the conventional saliency map based explanations.