«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

24

期数:

2014年03期

页码:

90-93

栏目:

智能、算法、系统工程

出版日期:

2014-03-31

文章信息/Info

Title:

Human Motion Capture Data Retrieval Based on Quaternion and EMD

文章编号:

1673-629X（2014）03-0090-04

作者:

肖秦琨¹; 李俊芳¹; 肖秦汉²

1.西安工业大学 电子信息工程学院；2.中国电子科技集团公司第20研究所军代室

Author(s):

XIAO Qin-kun[1]; LI Jun-fang[1]; XIAO Qin-han[2]

关键词:

运动捕获; 四元数; K-means聚类; EMD

Keywords:

motion capture; quaternion; K-means clustering; EMD

分类号:

TP391

文献标志码:

A

摘要:

针对运动捕获数据的高效匹配问题，提出了一种新的基于四元数描述和EMD( Earth Mover's Distance)的人体运动检索算法。该算法主要包括特征提取和运动匹配两部分。在特征提取部分，为了解决高维数据检索效率低的问题，引入了四元数描述符对关节点的数据信息特征进行描述，通过映射姿态分布的原始数据，并采取K-means聚类方法对待查询动作和运动数据库的特征数据进行降维并归类。在运动匹配部分，根据聚类结果，建立每个特征数据集的距离矩阵，将匹配问题转换为运输优化问题。然后，用EMD算法度量待查询动作和数据库动作之间的相似值。仿真实验结果证明了提出的算法是有效的。

Abstract:

A novel retrieval approach has been put forward based on quaternion and EMD ( Earth Mover's Distance) for efficient matc-hing motion capture data. The method mainly contains two steps:feature extracting and motion matching. In feature extracting part,for solving the problem of low efficiency retrieval of high-dimensional data,introduce the quaternion to represent key-joints rotation data in-formation,through the mapping original data of distribution,take K-means clustering methods to reduce the dimension and classify for the feature data in database of query motion and movement. In motion matching part,according to the clustering results,the distance matrix of each feature dataset was established,converting the matching problem of motion capture data into a optimization transportation problem. Then,the EMD measure algorithm was employed to query the similar value between query motions and database motions. Experiment re-sults show that the proposed approach is efficient.

常用功能

更新日期/Last Update: 1900-01-01