Human Action Recognition Using Spatio-Temporal Pyramid Model Based Background Subtraction on Depth Maps

Emmanuel MUTABAZI, Jian-jun NI, Ye YANG

Abstract


In this paper, a background subtraction in region method is proposed to recognize actions and interactions in the video. Firstly, the video is taken and converted into frames. Preprocessing techniques are applied to sampled images for noise reduction. Next, a background subtraction method is used to extract the foreground objects in region units. The combination of the background model, color of the object and movement information are employed to get the region object likelihood. Then, an Improved Adaptive region decision function determines the object regions. Moreover, the human detection method produces a bounding box surrounding a person. Histogram Oriented Gradient (HOG) is used for feature extraction and representation. Finally, Multi class support vector machine (SVM) is the classifier used for classification. Introduction

Keywords


Human Action Recognition, Background Subtraction in Regions, Improved Adaptive Region Decision Function, Histogram Oriented Gradient, Multi Class Support Vector Machine


DOI
10.12783/dtcse/cscbd2019/30077

Full Text:

PDF

Refbacks

  • There are currently no refbacks.