원문정보
초록
영어
A novel feature representation method based on AdaBoost algorithm is put forward for action recognition in this paper. The method can not only adequately describe action in complex scenarios, but also select the most discriminative sample subset from a large amount of raw features of training data. So it can realize a double result, that is, reduce the recognition computational complexity and achieve a good recognition accuracy. The pyramid histogram of oriented gradient feature (PHOG) descriptor is utilized to represent raw feature data. In order to select most discriminative samples subset, AdaBoost algorithm is used to extract the raw feature data. The nearest neighbor classifier algorithm is utilized to test the proposed method on the UCF Sports database. Experiment results show that the method not only achieve the better recognition rate but also greatly improve the speed of recognition.
목차
1. Introduction
2. Action Representation
2.1. The Raw Feature Extraction
2.2. AdaBoost for Raw Feature Selection
3. Classifier Design
4. Algorithm Verification and Results Analysis
4.1. The Raw Feature Extraction
4.2. Testing Result
5. Conclusion
Acknowledgments
References