ABSTRACT
The current state of art 1 allows achieving high action recognition accuracy but only after processing the entire video sequence, however for security issues, it is primordial to detect dangerous behaviors occurrence as soon as possible allowing early warnings. In this paper, we present a human activity recognition method using skeleton information provided by an RGB-D sensor by proposing a new descriptor modelling dynamic relation between 3D locations of skeleton joints, expressed in distance and spherical coordinates between the 20 available joints. First, we perform preprocessing step to recover missing skeleton information over frames, so, we normalize skeleton joints and apply a PCA dimension reduction to remove noisy information and enhance recognition accuracy while improving calculation and decision time. We also study the accuracy of the proposed descriptor calculated over limited few first frames to perform early action detection by specifying multiclass SVM classifier. We test this approach on two datasets, MSR Daily Activity 3D and our own dataset called INDACT; experimental evaluation shows that proposed approach can robustly classify actions outperforming state-of-art and maintain good accuracy score even using limited frame number.
- X. Yang and Y. Tian, 2014. Effective 3D action recognition using EigenJoints Journal of Visual Communication and Image Representation. Google ScholarDigital Library
- J Wang, Z Liu, Y Wu and J Yuan, 2012. Mining actionlet ensemble for action recognition with depth cameras, IEEE CVPR. Google ScholarCross Ref
- G. Cheng, Y. Wan, A. Saudagar, K. Numuduri and B. Buckles, 2015. Advances in human action recognition: A survey, Preprint arXiv: 1501.05964.Google Scholar
- T. Hassner. 2013. A Critical Review of Action Recognition Benchmarks, IEEE CVPR Workshops. Google ScholarDigital Library
- L Presti and M La Cascia, 2015. 3D Skeleton-based Human Action Classification: a Survey, pattern recognition journal.Google Scholar
- J. Shotton, A. Fitzgibbon, M. Cook, T. Sharp, M. Finocchio, R. Moore, A. Kipman, and A. Blake, 2011. Real-time human pose recognition in parts from single depth images IEEE CVPR.Google Scholar
- R. Gishick, J. Shotton, P. Kohli, A. Criminisi and A. Fitzgibbon, 2011. A Efficient regression of general activity human poses from depth images, IEEE ICCV.Google Scholar
- M Sun, P. Kohli and J. Shotton, 2012. Conditional regression forests for human pose estimation. IEEE CVPR.Google Scholar
- W. Li, Z. Zhang and Z. Liu, 2010. Action recognition based on a bag of 3D points, IEEE CVPR Workshop. Google ScholarCross Ref
- L Xia, C Chen and J Aggarwal, 2012. View invariant human action recognition using histograms of 3D joints, IEEE CVPR Workshop. Google ScholarCross Ref
- J Sung, C Ponce, B Selman and A Saxena, 2012. Unstructured human activity detection from RGBD images, IEEE ICRA.Google Scholar
- X Yang, C Zhang and Y Tian, 2012. Recognizing actions using depth motion maps based histograms of oriented gradients, ACM Multimedia. Google ScholarDigital Library
- H Wu, W Pan, X Xiong and S Xu, 2014. Human Activity Recognition Based on the Combined SVM&HMM, IEEE ICRA. Google ScholarCross Ref
- A Eweiwi, M.S Cheema, C Bauckhage and J Gall, 2014. Efficient pose-based action recognition, pattern recognition journal, ACCV.Google Scholar
- K Schindler.and L Gool, 2008. Action snippets: how many frames does human action recognition require? IEEE CVPR.Google Scholar
- J Chaquet, E Carmona.and A Fernandez-Caballero, 2013. A survey of video datasets for human action and activity recognition, Computer Vision and Image Understanding. Google ScholarDigital Library
- A Voulodimos, D Kosmopoulos, G Vasileiou and E Sardis, 2011. A dataset for workflow recognition in industrial scenes, IEEE ICIP. Google ScholarCross Ref
- A Shahroudy, J Liu, T Tsong and G Wang, 2016. NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis, IEEE CVPR.Google Scholar
Recommendations
Fusing Multiple Features for Depth-Based Action Recognition
Special Section on Visual Understanding with RGB-D SensorsHuman action recognition is a very active research topic in computer vision and pattern recognition. Recently, it has shown a great potential for human action recognition using the three-dimensional (3D) depth data captured by the emerging RGB-D ...
Evaluating spatiotemporal interest point features for depth-based action recognition
Human action recognition has lots of real-world applications, such as natural user interface, virtual reality, intelligent surveillance, and gaming. However, it is still a very challenging problem. In action recognition using the visible light videos, ...
Effective 3D action recognition using EigenJoints
HighlightsEffective method to recognize human actions using 3D skeleton joints.New action feature descriptor, EigenJoints, for action recognition.Accumulated Motion Energy (AME) method to perform informative frames selection.Our proposed approach ...
Comments