基于多视角自步学习的人体动作识别方法

计算机工程

基于多视角自步学习的人体动作识别方法

刘莹莹¹,邱崧 ¹,孙力 ¹,周梅 ¹,徐伟²

(1.华东师范大学信息科学技术学院上海市多维度信息处理重点实验室,上海 200241;2.上海交通大学图像处理与模式识别研究所,上海 200240)

收稿日期:2017-01-25 出版日期:2018-02-15 发布日期:2018-02-25
作者简介:刘莹莹(1992—),女,硕士研究生,主研方向为计算机视觉、机器学习;邱崧,讲师、博士;孙力,副教授、博士;周梅,讲师、博士;徐伟,博士。
基金资助:
国家自然科学基金(61302125,61377107);上海市自然科学基金(17ZR1408500)。

Human Action Recognition Method Based on Multi-view Self-paced Learning

LIU Yingying¹,QIU Song ¹,SUN Li¹,ZHOU Mei ¹,XU Wei ²

(1.Shanghai Key Laboratory of Multidimensional Information Processing,School of Information Science Technology,East China Normal University,Shanghai 200241,China;2.Institute of Image Processing and Pattern Recognition,Shanghai Jiaotong University,Shanghai 200240,China)

Received:2017-01-25 Online:2018-02-15 Published:2018-02-25

摘要/Abstract

摘要： 自步学习的动作识别方法采用课程学习的思路,忽略了不同视角动作特征对课程的影响,对多分类的人体两维视频复杂动作识别无法取得满意效果。针对上述问题,提出一种多视角自步学习算法。选取5个视角并提取Trajectory、HOG、HOF、MBHx和MBHy作为各自视角下的特征信息,利用自步学习算法学习得出对应视角下的动作分类课程,使用线性规划增强方法将不同视角下的课程进行融合,得出更适合解决多类复杂动作识别问题的综合课程。实验结果表明,相比单一视角自步学习方法和多视角支持向量机方法,该方法提高了多类复杂动作识别的效率和准确率,具有更高的可操作性和更广泛的应用前景。

关键词: 人体动作识别, 多视角融合, 自步学习, 线性规划增强, 支持向量机

Abstract: The action recognition method of step learning adopts the idea of curriculum learning,ignores the influence of different angles movement characteristics on the course,and can not achieve satisfactory results for the classification of two dimensional video complex action recognition.In order to solve the above problem,an algorithm for Multi-view Self-Paced Learning(MSPL) is proposed.It selects five views and extracts their features (Trajectory,HOG,HOF,MBHx and MBHy),and then learns curriculums under each view by Self-Paced Learning(SPL),fuses curriculums by means of Linear Programming Boosting(LPBoost),and learns a comprehensive curriculum that is more suitable for solving the problem of multi class complex action recognition at last.Experimental results show that compared with SPL and multi-view Support Vector Machine(SVM),the proposed algorithm improves the efficiency and accuracy of multi-class complex action recognition,and has higher operational and wider application prospects.

Key words: human action recognition, multi-view fusion, Self-Paced Learning(SPL), Linear Programming Boosting(LPBoost), Support Vector Machine(SVM)

中图分类号:

TP391

刘莹莹,邱崧,孙力,周梅,徐伟. 基于多视角自步学习的人体动作识别方法[J]. 计算机工程.

LIU Yingying,QIU Song,SUN Li,ZHOU Mei,XU Wei. Human Action Recognition Method Based on Multi-view Self-paced Learning[J]. Computer Engineering.

参考文献

参考文献［1］罗会兰,冯宇杰,孔繁胜.融合多姿势估计特征的动作识别［J］.中国图象图形学报,2015,20(11):1462-1472. ［2］LIU J,ALI S,SHAH M.Recognizing Human Actions Using Multiple Features［C］//Proceedings of IEEE Conference on Computer Vision and Pattern Recog-nition.Washington D.C.,USA:IEEE Press,2008:1-8. ［3］NEVATIA R,ZHAO T,HONGENG S.Hierarchical Language-based Representation of Events in Video Streams［C］//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition Workshop.Washington D.C.,USA:IEEE Press,2003:38-39. ［4］YAMATO J,OHYA J,ISHII K.Recognizing Human Action in Time-sequential Images Using Hidden Markov Model［C］//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,1992:379-385. ［5］SHI Q,CHENG L,WANG L,et al.Human Action Segmentation and Recognition Using Discriminative Semi-markov Models［J］.International Journal of Computer Vision,2011,93(1):22-32. ［6］朱国刚,曹林.基于Kinect传感器骨骼信息的人体动作识别［J］.计算机仿真,2014,31(12):329-333. ［7］NATARAIAN P,NEVATIA R.View and Scale Invariant Action Recognition Using Multiview Shape-flow Models［C］//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2008:1-8. ［8］BENGIO Y,LOURADOUR J,COLLOBERT R,et al.Curriculum Learning［C］//Proceedings of the 26th Annual International Conference on Machine Learning.New York,USA:ACM Press,2009:41-48. ［9］BENGIO Y,COURVILLE A,VINCENT P.Representation Learning:A Review and New Perspectives［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1798-1828. ［10］KUMAR M P,PACKER B,KOLLER D.Self-paced Learning for Latent Variable Models［C］//Proceedings of the 23th Annual Conference on Neural Information Processing Systems.Cambridge,USA:MIT Press,2010:1189-1197. ［11］JIANG L,MENG D,YU S I,et al.Self-paced Learning with Diversity［C］//Proceedings of the 27th Annual Conference on Neural Information Processing Systems.Cambridge,USA:MIT Press,2014:2078-2086. ［12］WANG H,KLASER A,SCHMID C,et al.Dense Trajectories and Motion Boundary Descriptors for Action Recognition［J］.International Journal of Computer Vision,2013,103(1):60-79. ［13］DALAL N,TRIGGS B.Histograms of Oriented Gradients for Human Detection［C］//Proceedings of IEEE Con-ference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2005:886-893. ［14］LAPTEV I,MARSZALEK M,SCHMID C.Learning Realistic Human Actions from Movies［C］//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2008:1-8. ［15］DALAL N,TRIGGA B,SCHMID C.Human Detection Using Oriented Histograms of Flow and Appearance［C］//Proceedings of European Conference on Computer Vision.Berlin,Germany:Springer,2006:428-444. ［16］GEHLER P V,NOWOZIN S.Let the Kernel Figure It out:Principled Learning of Preprocessing for Kernel Classi-fiers［C］//Proceedings of IEEE Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2009:2836-2843. ［17］许允喜,蒋云良,陈方.基于支持向量机增量学习和LPBoost的人体目标再识别算法［J］.光子学报,2011,40(5):758-763. ［18］方育柯,傅彦,周俊临,等.基于选择性集成的最大化软间隔算法［J］.软件学报,2013,34(5):1132-1147. ［19］XIA L,CHEN C C,AGGARWAL J K.View Invariant Human Action Recognition Using Histograms of 3D Joints［C］//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition Workshops.Washington D.C.,USA:IEEE Press,2012:20-27. ［20］SEIDENARI L,VARANO V,BERRETTI S,et al.Recognizing Actions from Depth Cameras as Weakly Aligned Multi-part Bag-of-poses［C］//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition Workshops.Washington D.C.,USA:IEEE Press,2013:479-485. ［21］汤荣志,段会川,孙海涛.SVM训练数据归一化研究［J］.山东师范大学学报(自然科学版),2016,31(4):60-65. 编辑吴云芳

[1]	江雨燕, 陶承凤, 李平. 数据增强和自适应自步学习的深度子空间聚类算法[J]. 计算机工程, 2023, 49(8): 96-103, 110.
[2]	马亚彤, 王松, 刘英芳. 融合多模态数据的人体动作识别方法研究[J]. 计算机工程, 2022, 48(9): 180-188.
[3]	王志江, 秦品乐, 柴锐, 武峰, 程一彤, 史玥. 基于深度学习的牙齿嵌塞自动判别方法[J]. 计算机工程, 2022, 48(4): 307-313.
[4]	王海, 翁晨傲, 李克, 骆曦. 一种面向基站扇区方向角估计的改进SVM算法[J]. 计算机工程, 2021, 47(4): 120-126.
[5]	张冰玉, 潘晴, 田妮莉, Everett Xiaolin Wang. 一种基于多重特征融合的信源个数估计方法[J]. 计算机工程, 2021, 47(4): 115-119,126.
[6]	连晓伟, 马垚, 陈永乐, 张壮壮, 王建华. 基于载荷特征与统计特征的Shodan流量识别[J]. 计算机工程, 2021, 47(1): 117-122.
[7]	袁哲明, 杨晶晶, 陈渊. 基于最大信息系数与冗余分摊的特征选择方法[J]. 计算机工程, 2020, 46(8): 101-105.
[8]	付子爔, 徐洋, 吴招娣, 许丹丹, 谢晓尧. 基于增量学习的SVM-KNN网络入侵检测方法[J]. 计算机工程, 2020, 46(4): 115-122.
[9]	张瑞, 陈红卫. 基于特征优化与SVPSO的工控入侵检测[J]. 计算机工程, 2020, 46(4): 19-25.
[10]	鲁淑霞, 蔡莲香, 张罗幻. 基于动量加速零阶减小方差的鲁棒支持向量机[J]. 计算机工程, 2020, 46(12): 88-95,104.
[11]	张波, 周从华, 张付全, 张婷, 蒋跃明. 一种面向SNP选择的模糊聚类算法[J]. 计算机工程, 2019, 45(8): 66-74.
[12]	周梦妮, 牛焱, 曹锐, 阎鹏飞, 相洁. 基于相位同步的癫痫信号识别与分析[J]. 计算机工程, 2019, 45(7): 291-295,302.
[13]	易国洪,代瑜,冯智莉,黎慧源. 基于SVM与DOM重心半径模型的Web正文提取[J]. 计算机工程, 2019, 45(6): 206-210.
[14]	苗续芝,陈伟,毕方明,房卫东,张武雄. 基于改进FOA-SVM的矿井火灾图像识别[J]. 计算机工程, 2019, 45(4): 267-274.
[15]	张裕平, 龚晓峰, 雒瑞森. 基于稀疏化双向二维主成分分析的人脸识别[J]. 计算机工程, 2019, 45(12): 232-236.

选择文件类型/文献管理软件名称

选择包含的内容