Head Pose Recognition with NNC-Trees

Abstract

Pose recognition is important in many practical applications. For example, a driver assistance system can detect if the driver is tired, sleepy, or careless from the poses. A pet robot can detect certain behavior patterns of the human user. The main purpose of this study is to develop a driver assistance system that can protect the drivers from careless accidents. As the first step, we propose a system for recognizing different poses of a human from the face images by using NNC-Tree. An NNC-Tree is a decision tree (DT) with each internal node containing a nearest neighbor classifier (NNC). We also developed a GUI for visualizing the prototypes in each NNC, as well as the whole tree. This interface makes it possible to understand, analyze, and reuse the learning results. This paper is a summary of what we have done so far.