当前位置: 首页 > 文章 > 一种基于混合特征的人体视角无关动作识别方法 信阳农林学院学报 2017 (2) 107-109
Position: Home > Articles > A Method of Human View-invariant Action Recognition Based on Mixed Feature Journal of Xinyang Agriculture and Forestry University 2017 (2) 107-109

一种基于混合特征的人体视角无关动作识别方法

作  者:
孙剑;肜丽
单  位:
信阳农林学院信息工程学院
关键词:
动作识别;视角无关;光流特征;兴趣点;混合特征
摘  要:
人体动作特征的选择直接关系到人体视角无关动作识别的准确率,单一的特征往往受到光照、环境、背景先验知识等条件的影响,识别效果不理想。本文提出了一种基于光流直方图与兴趣点视频段词袋直方图相结合的混合特征,该混合特征充分考虑了不同特征的优缺点,应用到视角无关的动作识别方法中。利用多视角IMAS动作识别数据库对该混合特征的实验结果表明,该特征对于人体视角无关动作识别具有较好识别效果。
译  名:
A Method of Human View-invariant Action Recognition Based on Mixed Feature
作  者:
SUN Jian;RONG Li;College of Information Engineering,Xinyang Agriculture and Forestry University;
关键词:
action recognition;;view-invariant;;optical flow;;interest points;;mixed feature
摘  要:
This paper proposed a mixed feature which combined the amplitude histogram of optical flow and the interest point words in shot length-based video,which fully considering the advantages and disadvantages of different characteristics. The experimental on multi-view action recognition dataset IMAS demonstrated that the proposed approach has satisfactory performance for the unknown view action recognition.

相似文章

计量
文章访问数: 7
HTML全文浏览量: 0
PDF下载量: 0

所属期刊

推荐期刊