On the Optimality of the Simple Bayesian Classifier under Zero-One Loss
Machine Learning - Special issue on learning with probabilistic representations
Robust Real-Time Periodic Motion Detection, Analysis, and Applications
IEEE Transactions on Pattern Analysis and Machine Intelligence
The Recognition of Human Movement Using Temporal Templates
IEEE Transactions on Pattern Analysis and Machine Intelligence
Space-Time Behavior Based Correlation
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Efficient Visual Event Detection Using Volumetric Features
ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1 - Volume 01
Multi-object image retrieval based on shape and topology
Image Communication
Behavior recognition via sparse spatio-temporal features
ICCCN '05 Proceedings of the 14th International Conference on Computer Communications and Networks
Towards optimal bag-of-features for object categorization and semantic video retrieval
Proceedings of the 6th ACM international conference on Image and video retrieval
Tracking and recognizing actions of multiple hockey players using the boosted particle filter
Image and Vision Computing
Towards Real-Time Human Action Recognition
IbPRIA '09 Proceedings of the 4th Iberian Conference on Pattern Recognition and Image Analysis
Hi-index | 0.02 |
In this work, a schematic model for human activity recognition based on multiple cues is introduced. In the beginning, a sequence of temporal silhouettes of the moving human body parts are extracted from a video clip (i.e., an action snippet). Next, each action snippet is temporally split into several time-slices represented by fuzzy intervals. As shape features, a variety of descriptors both boundary-based (Fourier descriptors, Curvature features) and region-based (Moments, Momentbased features) are then extracted from the silhouettes at each time-slice. Finally, an NB (Naïve Bayes) classifier is learned in the feature space for activity classification. The performance of the method was evaluated on the KTH dataset and the obtained results are quite encouraging and show that an accuracy on par with or exceeding that of existing methods is achievable. Further the simplicity and computational efficiency of the features employed allow the method to achieve real-time performance, and thus it can provide latency guarantees to real-time applications.