Profile-based Human Action Recognition using Depth Information

Pongsatorn Chawalitsittikul and Nikom Suvonvorn

Keywords

Action recognition, Video surveillance, Human modeling, Multiviews, RGB D image

Abstract

The recognition of human actions is an important step for human behaviour understanding via video processing. In this paper, we propose an accurate method of modelbased action recognition using color and depth information. The parametric model of human is extracted from image sequences using mixture of motion, color and depth information. Extracted model features are then classified into five basic actions using artificial neural network. The experimentation result shows that our method gives high recognition rate more than 90%.

Important Links:



Go Back