您的位置:首页  > 论文页面

基于时空纹理特征的三维人体行为分析系统

发表时间:2015-02-28  浏览量:2771  下载量:1310
全部作者: 王光超,范春晓
作者单位: 北京邮电大学电子工程学院
摘 要: 当前面向深度视频信息的三维(three-dimensional,3D)人体行为视频被广泛应用于体感交互和行为分析领域中,为在现实场景中更好地分析人体行为,提出一种基于时空纹理特征的人体行为视频识别方法和系统。首先利用Kinect等体感设备采集人体行为视频,并对行为视频进行预处理,然后对视频进行3D时空纹理特征提取,生成人体行为特征信息模型,进而完成分类识别;最后在行为分析系统上实现模拟个性化控制。相比于过去的识别方法,文中方法简化了视频人体行为识别的复杂度,降低了行为特征数据量,并取得了较好的系统分析效果。
关 键 词: 模式识别;计算机视觉;人体行为识别;时空纹理特征;深度视频;行为分析系统
Title: Three-dimensional human behavior analysis system based on spatiotemporal texture feature
Author: WANG Guangchao, FAN Chunxiao
Organization: School of Electronic Engineering, Beijing University of Posts and Telecommunications
Abstract: Current three-dimensional (3D) human behavior videos containing depth information have been widely applied to areas of body sensing interaction and behavior analysis. In order to better analyze human behavior in real-life scenarios, a human behavior video recognition method and system is proposed based on temporal texture features. First of all, Kinect and other somatosensory equipment are employed to capture human behavior videos, which are then preprocessed. Afterwards, the 3D temporal texture features are extracted to generate a human behavior feature information model. Then classification and recognition is accomplished and analog personality control is realized on the behavior analysis system. Compared to previous recognition methods, the proposed method has simplified the complexity of human behavior recognition in the videos, reduced the data size of behavioral features, and achieved good system analysis results.
Key words: pattern recognition; computer vision; human behavior recognition; spatiotemporal texture features; depth videos; behavior analysis system
发表期数: 2015年2月第4期
引用格式: 王光超,范春晓. 基于时空纹理特征的三维人体行为分析系统[J]. 中国科技论文在线精品论文,2015,8(4):316-324.
 
0 评论数 0
暂无评论
友情链接