QQ客服 官方微博 反馈留言

	Unsupervised Human Motion Analysis Using Automatic Label Tree
	Kui Jia; Xie Wuyuan
	2009
会议名称	2009 IEEE International Conference on Systems, Man and Cybernetics, SMC 2009
会议地点	Hong Kong
英文摘要	Giving different human motions, there may be similar motion features in some body components, e.g., the motion features of arms when performing jogging and running, which are thus less discriminative for motion classification. In this paper, we consider counting less on these body components that have less discriminative information amongst different human motions. To this end, we present a new topic model, probabilistic latent semantic analysis based on multiple bags (M-PLSA), in which not all body components are considered equally important, i.e.,motion features of less discriminative components are made less use of so that their contributions for classification are reduced. We use sparse spatio-temporal features extracted from videos to create visual words which are later assigned to different body components that they are detected from, so that co-occurrence matrices of different components can be calculated based on their corresponding vocabularies. Such label task can beautomatically fulfilled by using the query visual words, i.e., words whose component labels are unknown, to traverse an automatic label tree (ALT) that grows from the training words with component labels. We show the performance of our approach on KTH dataset
收录类别	EI
语种	英语
内容类型	会议论文
源URL	[http://ir.siat.ac.cn:8080/handle/172644/2457]
专题	深圳先进技术研究院_集成所
作者单位	2009
推荐引用方式 GB/T 7714	Kui Jia,Xie Wuyuan. Unsupervised Human Motion Analysis Using Automatic Label Tree[C]. 见:2009 IEEE International Conference on Systems, Man and Cybernetics, SMC 2009. Hong Kong.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。