Motion description with local binary pattern and motion history image: Application to human motion recognition

Abstract

This paper presents a new human motion recognition method based on motion history image (MHI) and local binary pattern (LBP). MHI describes human motion sequence in one gray level image and LBP extracts its texture features. LBP feature image is built and chi square distance is applied to compute matching cost. Experiments are conducted with encouraging results which show a success of applying LBP in motion recognition.