Fusion of Color and Depth Video for Human Behavior Recognition in an Assistive Environment

Dimitrios I. Kosmopoulos, Paul Doliotis, Vassilis Athitsos, and Ilias Maglogiannis


Abstract:In this paper we investigate the effects of fusing feature streams extracted from color and depth videos, aiming to monitor the actions of people in an assistive environment. The output of fused time- series classifiers is used to model and extract actions. To this end we compare the Hidden Markov model classifier and fusion methods like early, late or state fusion. Our experiments employ a public dataset, which was acquired indoors.