Rectangular layer model for profile-based human action recognition using multi view depth information

Pongsagorn Chalearnnetkul

Department of Computer Engineering, Faculty of Engineering, Prince of Songkla University, Hat Yai, Songkhla, Thailand

Nikom Suvonvorn

Department of Computer Engineering, Faculty of Engineering, Prince of Songkla University, Hat Yai, Songkhla, Thailand

DOI: https://doi.org/10.14456/apst.2017.23

Abstract

Human action recognition is a fundamental component for understanding complex activities or behaviors, especially for video surveillance and health-care applications. In this paper we introduce profile-based human action recognition from multi-view cameras using RGB-D information through a Rectangular Layer Model (RLM). Our model tended to show improved performance when the perspective distortion or the lack of information occurred during a single-view approach. A fusion model was tested for five basic actions: walking and standing, sitting, bending, and laying, and at different perspective viewpoints. The system can perform at 28.99 fps while its overall precision is about 92.25%.

How to Cite

Chalearnnetkul, P., & Suvonvorn, N. (2017). Rectangular layer model for profile-based human action recognition using multi view depth information. Asia-Pacific Journal of Science and Technology, 22(3), APST–22. https://doi.org/10.14456/apst.2017.23

References

Liu J, Shah M, Kuipers B, Savarese S. Cross-view action recognition via view knowledge transfer. Proc Computer Vision Pattern Recognition IEEE. 2011;3209-3216.

Gkalelis N, Nikolaidis N, Pitas, I. View independent human movement recognition from multi-view video exploiting a circular invariant posture representation. Proc Multi Media and Expo IEEE. 2009;394-397.

Souvenir R, Babbs J. Learning the viewpoint manifold for action recognition. Proc Computer Vision Pattern Recognition IEEE. 2008;1-7.

Ahmad M, Lee SW. Hmm-based human action recognition using multi-view image sequences. Proc Pattern Recognition IEEE. 2006;1:263-266.

Tran C, Trivedi MM. Human body modeling and tracking using volumetric representation Selected recent studies and possibilities for extensions. Proc Distributed Smart Cameras ACM/IEEE. 2008;1-9.

Pehlivan S, Duyugulu P. A new pose-based representation for recognizing actions from multiple cameras. Journal of Computer Vision and Image Understanding ACM. 2010;115(2):140–151.

Holte MB, Moeslund TB, Nikolaidis N, Pitas I. 3D human action recognition for multi-view camera systems. Proc 3D Imaging, Modeling, Processing, Visualization and Transmission IEEE. 2011;342-349.

Huang P, Hilton A, Starck J. Shape similarity for 3D video sequences of people. Journal of Computer Vision Springer. 2010;89(2):362-381.

Chawalitsittikul P, Suvonvorn N. Profile-based Human Action Recognition using Depth Information. Proc Advances Computer Science and Engineering ACTA Press. 2012;376–380.

Noorit N, Suvonvorn N , Karnchanadecha M. Model-based Human Action Recognition. Proc Digital Image Processing SPIE. 2010;7546.

Chuang CH, Hsieh JW, Tsai LW, Fan KC. Human Action Recognition Using Star Templates and Delaunay Triangulation. Proc Intelligent Information Hiding and Multimedia Signal Processing. 2008;179-182.

Parisi GI, Weber C, Wermter S. Human action recognition with hierarchical growing neural gas learning. Proc Artificial Neural Networks Springer. 2014;8681:89–96.

Sawant N, Biswas KK. Human Action Recognition Based on Spatio-temporal Features. Proc Pattern Recognition and Machine Intelligence Springer. 2009;5909:357-362.

KaewTraKuPong P, Bowden R. An improved adaptive background mixture model for real-time tracking with shadow detection. Proc Advanced Video-Based Surveillance Systems Springer, 2001;135-144.

Chawalitsittikul P. Side-View Based Human Action Recognition Using Stereo Vision [Thesis]. Songkhla: Prince of Songkla University; 2013, Thai.