An Attention Selection Model with Visual Memory and Online Learning

Abstract

In this paper, an attention selection model with visual memory and online learning is proposed, which has three parts: Sensory Mapping (SM), Cognitive Mapping (CM) and Motor Mapping (MM). CM is the novelty of our model which incorporates visual memory and online learning. In order to mimic visual memory, we put forward an Amnesic Incremental Hierachical Discriminant Regression (AIHDR) Tree which has an amnesic function to guide the deletion of redundant information of the tree. Experimental results show that our AIHDR tree has better performance in retrieval speed and accuracy than IHDR/HDR tree. Self-Supervised Competition Neural Network (SSCNN) in CM has the characteristics of online learning since its connection weights can be updated in real time according to the change of environment. Eyeball Movement Prediction (EMP) mechanism is applied to estimate the movement of human eyeball so that attention can be focused on interested objects. Several applications such as object tracking and robot self-localization are realized by our proposed work.