Title: Human action recognition in videos using structure similarity of aligned motion images

Authors: Salim Al-Ali; Mariofonna Milanova

Addresses: Department of Applied Science, University of Arkansas at Little Rock, Little Rock, AR 72204, USA; Dohuk Polytechnic University, Duhok, Iraq ' Department of Computer Science, University of Arkansas at Little Rock, Little Rock, AR 72204, USA

Abstract: Human action recognition in videos is a prominent field in image processing research and it is a frequently required technique for many computer vision applications. In this paper, a new novel algorithm for human action recognition is presented. This algorithm depends on aligned motion images (AMIs). Three types of AMIs are used: aligned motion history image (AMHI), aligned motion energy image (AMEI), and aligned gait energy image (AGEI). As a feature, these AMIs are obtained after frame processing, AMI computing, and AMI processing. In order to classify and identify the activity, structure similarity measurement index is applied for first time in human recognition in this paper. The achieved experimental results demonstrate a high level of accuracy and efficiency. These results are very close to the recognition of the human observer since accomplished result is approximately 98.93% by using the AGEI that calculated from a sub sequence of first 33 frames in each video samples.

Keywords: human activity recognition; motion history images; motion energy images; gait energy images; aligned motion images; structure similarity index measurement; human action; videos; image processing; computer vision.

DOI: 10.1504/IJRIS.2014.063945

International Journal of Reasoning-based Intelligent Systems, 2014 Vol.6 No.1/2, pp.71 - 82

Received: 23 Jul 2013
Accepted: 24 Dec 2013

Published online: 27 Jul 2014 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article