Human Classification for A Partner Robot Based on Self Organising Maps and Back-Propagation

Abstract

Basically, a robot should perform moving object extraction, clustering, and classification for visual perception used in the interaction with human. For this, we will propose a total system for human classification for a partner robot. The differential filter will be used for detecting a moving object. Next, the color combination pattern will be extracted by k-means. A robot will classify the detected human by self-organizing map based on the color pattern and a back-propagation of neural network will be used at the end of process.