Classified Judgment on Same-different Based on Bayesian Inference

Abstract

Abstract-The ability to detect objects precisely with the same orientations from random orientations is one of the most remarkable functions of perception and cognition. Many psychophysical evidences imply human's behavior follows near-optimal Bayesian inference in tasks of decision making. We explore the model of bars with the same orientation from the random orientations by the means of theories and experiments. In experiments, the numbers and locations of bars with same orientation are randomly assigned. The performances of subjects depend on the number of bars (set size), the number of bars with same orientations and reliabilities of measurements. We put forward a model based on Bayesian theory in visual search to judge how many bars with same orientation among a mass of orientations. To compare the effect of the model, we design two alternative models. We find our model on this kind of classified judgment matches well the experimental data, is superior to alternative models, and it provides a normative and mathematically quantitative description.