Action recognition via cumulative histogram of multiple features

Xunshi Yan; Yupin Luo

doi:10.1117/1.3530070

1 January 2011 Action recognition via cumulative histogram of multiple features

Xunshi Yan, Yupin Luo

Author Affiliations +

Optical Engineering, Vol. 50, Issue 1, 017203 (January 2011). https://doi.org/10.1117/1.3530070

Abstract

Spatial-temporal interest points (STIPs) are popular in human action recognition. However, they suffer from difficulties in determining size of codebook and losing much information during forming histograms. In this paper, spatial-temporal interest regions (STIRs) are proposed, which are based on STIPs and are capable of marking the locations of the most "shining" human body parts. In order to represent human actions, the proposed approach takes great advantages of multiple features, including STIRs, pyramid histogram of oriented gradients and pyramid histogram of oriented optical flows. To achieve this, cumulative histogram is used to integrate dynamic information in sequences and to form feature vectors. Furthermore, the widely used nearest neighbor and AdaBoost methods are employed as classification algorithms. Experiments on public datasets KTH, Weizmann and UCF sports show that the proposed approach achieves effective and robust results.

©(2011) Society of Photo-Optical Instrumentation Engineers (SPIE)

Citation Download Citation

Xunshi Yan and Yupin Luo "Action recognition via cumulative histogram of multiple features," Optical Engineering 50(1), 017203 (1 January 2011). https://doi.org/10.1117/1.3530070

Published: 1 January 2011

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available