Dynamic Feature Selection for Online Action Recognition

The ability to recognize human actions in real-time is fundamental in a wide range of applications from home entertainment to medical systems. Previous work on online action recognition has shown a tradeoff between accuracy and latency. In this paper we present a novel algorithm for online action recognition that combines the discriminative power of Random Forests for feature selection and a new dynamic variation of AdaBoost for online classification. The proposed method has been evaluated using datasets and performance metrics specifically designed for real time action recognition. Our results show that the presented algorithm is able to improve recognition rates at low latency.

[1]  Serge J. Belongie,et al.  Behavior recognition via sparse spatio-temporal features , 2005, 2005 IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance.

[2]  Wei-Yin Loh,et al.  Classification and regression trees , 2011, WIREs Data Mining Knowl. Discov..

[3]  Ron Kohavi,et al.  Wrappers for Feature Subset Selection , 1997, Artif. Intell..

[4]  J. Friedman.  Special Invited Paper-Additive logistic regression: A statistical view of boosting , 2000 .

[5]  Leo Breiman,et al.  Classification and Regression Trees , 1984 .

[6]  Hans-Paul Schwefel,et al.  Advances in Computational Intelligence , 2003, Natural Computing Series.

[7]  Helena M. Mentis,et al.  Instructing people for training gestural interactive systems , 2012, CHI.

[8]  J.K. Aggarwal,et al.  Human activity analysis , 2011, ACM Comput. Surv..

[9]  Alexandros André Chaaraoui,et al.  Optimal Joint Selection for Skeletal Data from RGB-D Devices Using a Genetic Algorithm , 2012, MICAI.

[10]  Barbara Caputo,et al.  Recognizing human actions: a local SVM approach , 2004, ICPR 2004.

[11]  James W. Davis,et al.  The Recognition of Human Movement Using Temporal Templates , 2001, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[13]  Joseph J. LaViola,et al.  Exploring the Trade-off Between Accuracy and Observational Latency in Action Recognition , 2013, International Journal of Computer Vision.

[14]  Sebastian Nowozin,et al.  Action Points: A Representation for Low-latency Online Human Action Recognition , 2012 .

[15]  Junji Yamato,et al.  Recognizing human action in time-sequential images using hidden Markov model , 1992, Proceedings 1992 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[16]  Jake K. Aggarwal,et al.  Human action recognition with extremities as semantic posture representation , 2009, 2009 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops.

[17]  Xin Miao,et al.  A comparison of random forest and Adaboost tree in ecosystem classification in east Mojave Desert , 2010, 2010 18th International Conference on Geoinformatics.

[18]  Isabelle Guyon,et al.  An Introduction to Variable and Feature Selection , 2003, J. Mach. Learn. Res..

[19]  Ryan M. Rifkin,et al.  In Defense of One-Vs-All Classification , 2004, J. Mach. Learn. Res..

[20]  Sriram Subramanian,et al.  Talking about tactile experiences , 2013, CHI.

[21]  Mario Cannataro,et al.  Protein-to-protein interactions: Technologies, databases, and algorithms , 2010, CSUR.

[22]  Edoardo Amaldi,et al.  On the Approximability of Minimizing Nonzero Variables or Unsatisfied Relations in Linear Systems , 1998, Theor. Comput. Sci..

[23]  Dimitrios Makris,et al.  G3D: A gaming action dataset and real time action recognition evaluation framework , 2012, 2012 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops.

[24]  Yoav Freund,et al.  Experiments with a New Boosting Algorithm , 1996, ICML.