Human Tracking Based on Hierarchical Attention Control

A hierarchical attention control model is introduced for implementing a human-like attention control in computer vision. The lowest layer of the hierarchy is called a window layer, in which each attention window is controlled by signal processing in the same manner as in the conventional attention control. The other four layers, however, are designed for implementing an attention control with high-level information processings. In our model, the attention control is accomplished using a tree structure across the five layers and both the bottom-up and top-down processings are accomplished on it. This paper provides its application to human tracking in complex situations with successful experimental results.