An Invariant Pattern Recognition System Using the Bayesian Inference on Hierarchical Sequences with Pre-processing

Abstract

The human being can understand real-world objects based on some kinds of invariable characteristics. Recently, a mathematical model for this has been proposed that is based on the Bayesian inference on hierarchical sequences by George, D. and Hawkins, J. (2004). It assumed that human brain cortex solves the invariance problem in a manner that is using a multi-hierarchical structure. When we applied the model to a line Drawing Recognition System (DRS), however, the performance was not as good as we had expected. This is especially the case when the hand input character is too small or too big. In this paper, we propose a method for improving this. Our method is based on a fact that human eyes are able to automatically focus on the object by its position, size, and lightness. That is, before the recognition, we perform a piece of pre-processing so that it can adjust the position, size and the lightness to make them most suitable for the recognition followed.