[1]徐濛,彭淑娟,柳欣.结合金字塔模型和随机森林的运动捕获序列语义标注[J].华侨大学学报(自然科学版),2017,38(6):848-853.[doi:10.11830/ISSN.1000-5013.201601011]
　XU Meng,PENG Shujuan,LIU Xin.Motion Capture Sequence Semantic Annotation Via Pyramid Model and Random Forests[J].Journal of Huaqiao University(Natural Science),2017,38(6):848-853.[doi:10.11830/ISSN.1000-5013.201601011]
点击复制

结合金字塔模型和随机森林的运动捕获序列语义标注()
分享到：

《华侨大学学报（自然科学版）》[ISSN:1000-5013/CN:35-1079/N]

卷:
第38卷
期数:
2017年第6期
页码:
848-853
栏目:
出版日期:
2017-11-20

文章信息/Info

Title:
Motion Capture Sequence Semantic Annotation Via Pyramid Model and Random Forests
文章编号:
1000-5013(2017)06-0848-06
作者:
徐濛;  彭淑娟;  柳欣
华侨大学 计算机科学与技术学院, 福建 厦门 361021
Author(s):
XU Meng;  PENG Shujuan;  LIU Xin
College of Computer Science and Technology, Huaqiao University, Xiamen 361021, China
关键词:
语义标注;  概率主成分分析;  傅里叶时间金字塔;  随机森林
Keywords:
semantic annotation;  probabilistic principal component analysis;  fourier temporal pyramid;  random forests
分类号:
TP391
DOI:
10.11830/ISSN.1000-5013.201601011
文献标志码:
A
摘要:
针对原始运动捕获数据结构复杂、语义模糊的问题,提出一种结合金字塔模型和随机森林的运动捕获序列语义标注方法.首先,利用概率主成分分析将运动序列划分为具有特定语义的运动片段.然后,将运动片段的欧拉角数据转换为人体各个关节点的三维空间位置坐标数据,统一骨骼长度,提取运动数据的2种互补性几何特征,并分别归一化.再次,运用傅里叶时间金字塔模型构建运动片段完整的时空特征.最后,利用已训练的随机森林分类器对各个运动片段进行标注.结果表明:该方法能够对具有不同语义的复杂运动序列进行有效标注,且可用于不同表演者,具有一定的实用性和通用性.
Abstract:
According to the complexity and semantic ambiguity within original motion capture data, we presents an effective motion capture sequence semantic approach via pyramid model and random forests. Firstly, utilize probabilistic principal component analysis to segment motion sequences into several motion clips with certain semantics. Then, the Euler angler data of each motion clip are transformed into three-dimensional space coordinates of each human joint, and the bone lengths are unified. Subsequently, two complementary features are extracted and normalized. Accordingly, the Fourier temporal pyramid model is adopted to represent the spatiotemporal characteristics of motion clips. Finally, the trained random forests classifier is employed to label each motion clip. The proposed approach is able to well annotate complex motion sequences effectively and can be applied to different performers. The experimental results show that it has certain practically and generality.

参考文献/References:

[1] ADISTAMBHA K,RITZ C H,BURNETT I S.Motion classification using dynamic time warping[C]//Proceedings of the Multimedia Signal Workshops.Cairns:IEEE Press,2008:622-627.
[2] LYU Fengjun,NEVATIA R.Recognition and segmentation of 3D human action using HMM and multi-class adaboost[C]//Proceedings of the European Conference on Computer Vision.Graz:Springer Press,2006:359-372.
[3] ZHU Hongli,DU Pengying,XIANG Jian.3D Motion Recognition based on ensemble learning[C]//Proceedings of the International Conference on Image Analysis for Multimedia Interactive Services Workshops.Santorini:IEEE Press,2007:1-4.
[4] XIA Lu, CHEN C C, AGGARWAL J K. View invariant human action recognition using histograms of 3D joints[C]//Proceedings of the International Conference on Computer Vision and Pattern Recognition.Providence:IEEE Press,2012:20-27.DOI:10.1109/CVPRW.2012.6239233.
[5] BENGALUR M D.Human activity recognition using body pose features and support vector machine[C]//Proceedings of the International Conference on Advances in Computing, Communications and Informatics.Mysore:IEEE Press,2013:1970-1975.DOI:10.1109/ICACCI.2013.6637484.
[6] HAN Lei,WEI Liang,WU Xinxiao,et al.Human action recognition using discriminative models in the learned hierarchical manifold space[C]//Proceedings of the International Conference on Automatic Face and Gesture Recognition.Amsterdam:IEEE Press,2008:1-6.
[7] LI Chuanjun,ZHENG Soqing,PRABHAKARAN B.Segmentation and recognition of motion streams by similarity search[J].ACM Transactions on Multimedia Computing, Communications, and Applications,2007,3(3):79-82.DOI:10.1145/1236471.1236475·Source:DBLP.
[8] MULLER M,ROER T,CLAUSEN M,et al.Documentation: Mocap da-tabase HDM05[EB/OL].[2016-03-02] .http://www.mpi-inf.mpg.de/resources/HDM05.
[9] BARBI V C J,SAFONOVA A,PAN J,et al.Segmenting motion capture data into distinct behaviors[C]//Proceedings of the International Conference on Graphics Interface.London:Canadian Human-Computer Communications Society,2004:185-194.
[10] SHUM H,HO E S.Real-time physical modelling of character movements with Microsoft kinect[C]//Proceedings of the Symposium on Virtual Reality Software and Technology.Toronto:ACM Press,2012:17-24.
[11] 杨跃东,王莉莉,郝爱民,等.基于几何特征的人体运动捕获数据分割方法[J].系统仿真学报,2007,19(10):2229-2234.DOI:10.3969/j.issn.1004-731X.2007.10.022.
[12] 彭淑娟.基于中心距离特征的人体运动序列关键帧提取[J].系统仿真学报,2012,24(3):565-569.
[13] WANG Jiang,LIU Zicheng,WU Ying,et al.Learning actionlet ensemble for 3D human action recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,36(5):914-927.
[14] BREIMAN L.Random forests [J].Machine Learning,2001,45(1):5-32.

相似文献/References:

[1]乔虎,周源,白瑀.采用特征识别技术的MBD模型自动语义标注方法[J].华侨大学学报(自然科学版),2018,39(5):750.[doi:10.11830/ISSN.1000-5013.201804086]
　QIAO Hu,ZHOU Yuan,BAI Yu.Automatic Semantic Tagging of MBD Model Using Feature Recognition[J].Journal of Huaqiao University(Natural Science),2018,39(6):750.[doi:10.11830/ISSN.1000-5013.201804086]

备注/Memo

备注/Memo:
收稿日期: 2015-01-06
通信作者: 彭淑娟(1982-),女,讲师,博士,主要从事计算机视觉与计算机动画的研究.E-mail:pshujuan@hqu.edu.cn.
基金项目: 国家自然科学基金资助项目(61202298, 61300138); 福建省自然科学基金资助项目(2014J01239, 2015J01656); 华侨大学高层次人才科研启动项目(14BS207); 华侨大学中青年科研提升计划(ZQN-PY309); 华侨大学研究生科研创新能力培育计划资助项目(1400414009)
更新日期/Last Update: 2017-11-20