摘要
This paper proposes a framework for human action recognition based on procrustes analysis and Fisher vector coding(FVC).Firstly,we applied a pose feature extracted from silhouette image by employing Procrustes analysis and local preserving projection(LPP).Secondly,the extracted feature can preserve the discriminative shape information and local manifold structure of human pose and is invariant to translation,rotation and scaling.Finally,after the pose feature was extracted,a recognition framework based on FVC and multi-class supporting vector machine was employed to classify the human action.Experimental results on benchmarks demonstrate the effectiveness of the proposed method.
This paper proposes a framework for human action recognition based on procrustes analysis and Fisher vector coding(FVC). Firstly, we applied a pose feature extracted from silhouette image by employing Procrustes analysis and local preserving projection(LPP). Secondly, the extracted feature can preserve the discriminative shape information and local manifold structure of human pose and is invariant to translation, rotation and scaling. Finally, after the pose feature was extracted, a recognition framework based on FVC and multi-class supporting vector machine was employed to classify the human action. Experimental results on benchmarks demonstrate the effectiveness of the proposed method.
作者
CAI Jiaxin
ZHONG Ranxu
LI Junjie
蔡加欣;钟然旭;李俊杰(School of Applied Mathematics,Xiamen University of Technology,Xiamen 361024,China;Department of Software Research and Development,Guangdong Grandmark Automotive Systems Co.,Ltd.,Dongguan 523000,China)
基金
National Natural Science Foundation of China(No.61602148)
Natural Science Foundation of Fujian Province,China(No.2016J01040)
Xiamen University of Technology High Level Talents Project,China(No.YKJ15018R)
