基于函数型数据分析方法的人体动态行为识别

苏本跃; 蒋京; 汤庆丰; 盛敏

doi:10.16383/j.aas.2017.c160120

基于函数型数据分析方法的人体动态行为识别

doi: 10.16383/j.aas.2017.c160120

苏本跃^1,2, ,,
蒋京^1,2,,
汤庆丰^1,2,,
盛敏^2,3,

1.
安庆师范大学计算机与信息学院安庆 246133
2.
智能感知与计算安徽省高校重点实验室安庆 246133
3.
安庆师范大学数学与计算科学学院安庆 246133

基金项目:

安徽省高校自然科学研究重点项目 KJ2014A142

国家自然科学基金 11471093

情感计算与先进智能机器安徽省重点实验室开放课题 ACAIM160102

国家科技支撑课题 2014BAH13F02

详细信息

作者简介:
蒋京  安庆师范大学计算机与信息学院, 智能感知与计算安徽省高校重点实验室硕士研究生.主要研究方向为机器学习与模式识别.E-mail:jjaq1990@sina.com

汤庆丰  安庆师范大学计算机与信息学院, 智能感知与计算安徽省高校重点实验室硕士研究生.主要研究方向为机器学习与模式识别.E-mail:qftang1991@sina.com

盛敏安  庆师范大学数学与计算科学学院, 智能感知与计算安徽省高校重点实验室副教授.2009年获得合肥工业大学计算机与信息学院博士学位.主要研究方向为模式识别与图像及视频处理.E-mail:msheng0125@aliyun.com

通讯作者:
苏本跃安庆师范大学计算机与信息学院, 智能感知与计算安徽省高校重点实验室教授.2007年获得合肥工业大学计算机与信息学院博士学位.主要研究方向为模式识别与机器学习, 图形图像处理.E-mail:bysu@aqnu.edu.cn

计量
- 文章访问数: 2389
- HTML全文浏览量: 500
- PDF下载量: 1227
- 被引次数: 0
出版历程
- 收稿日期: 2016-02-18
- 录用日期: 2016-07-18
- 刊出日期: 2017-05-01

Human Dynamic Action Recognition Based on Functional Data Analysis

SU Ben-Yue^{1,2
, ,},
JIANG Jing^{1,2
,},
TANG Qing-Feng^{1,2
,},
SHENG Min^{2,3
,}

1.
School of Computer and Information, Anqing Normal University, Anqing 246133
2.
University Key Laboratory of Intelligent Perception and Computing of Anhui Province, Anqing 246133
3.
School of Mathematics and Computational Science, Anqing Normal University, Anqing 246133

Funds:

Natural Science Research Funds of Education Department of Anhui Province KJ2014A142

National Natural Science Foundation of China 11471093

Anhui Province Key Laboratory of Affective Computing and Advanced Intelligent Machine ACAIM160102

National Science and Technology Support Program 2014BAH13F02

More Information

Author Bio:
Master student at the School of Computer and Information, Anqing Normal University and the University Key Laboratory of Intelligent Perception and Computing of Anhui Province. His research interest covers machine learning and pattern recognition

Master student at the School of Computer and Information, Anqing Normal University and the University Key Laboratory of Intelligent Perception and Computing of Anhui Province. His research interest covers machine learning and pattern recognition

Associate professor at the School of Mathematics and Computational Science, Anqing Normal University and the University Key Laboratory of Intelligent Perception and Computing of Anhui Province. She received her Ph.D. degree from the School of Computer and Information, Hefei University of Technology in 2009. Her research interest covers pattern recognition and image and video processing

Corresponding author: SU Ben-Yue Professor at the School of Computer and Information, Anqing Normal University and the University Key Laboratory of Intelligent Perception and Computing of Anhui Province. He received his Ph.D. degree from the School of Computer and Information, Hefei University of Technology in 2007. His research interest covers pattern recognition and machine learning, image processing and computer graphics. Corresponding author of this paper

摘要

摘要: 人体运动过程中，肢体的运动是连续的，而对应的运动捕捉数据是离散的.为了更好地分析人体日常运动行为的连续性与周期性，本文提出了一种基于函数型数据分析（Functional data analysis，FDA）的人体动态行为识别方法.首先，利用函数型数据分析方法，将可穿戴式运动捕捉系统采集的人体周期行为数据函数化，通过函数准确地定义数据的连续性与周期性；然后，根据导函数信息确定一个运动周期的起始点，并近似地提取出一个运动周期的数据序列；最后，根据不同行为一个周期内的曲线特征差异，利用支持向量机对动态行为进行分类识别.实验结果表明，本文的算法既能够较好地描述人体动态行为的连续性与周期性，又使得运动数据在标定的统一起始点处对齐，且在WARD数据集与自采集数据集上均取得了较好的识别率，分别达到97.5%与98.75%.
- 动态行为识别 /
- 连续性与周期性 /
- 周期行为 /
- 函数型数据分析 /
- 可穿戴式运动捕捉系统
Abstract: In human motion, limb movement is continuous. However, the corresponding motion capture data is discrete. This paper explores a method for human dynamic action recognition based on functional data analysis (FDA) so as to analyze the continuity and periodicity of daily action. Firstly, we transform the periodic data collected by the wearable motion capture system into functional data using FDA, and then define the continuity and periodicity of data exactly by using function properties. Secondly, we determine the initial point of a motion period according to the derivative information, and then extract the data series representing a period of motion. Finally, we utilize support vector machine (SVM) to classify the dynamic action according to the different characteristics of the curves about different actions in a period. The experimental result indicates that our algorithm can describe the continuity and periodicity of human dynamic action, and align the motion data at the uniform start point we determined. At the same time, desirable recognition rates, such as 97.5% and 98.75%, can be achieved based on WARD and our database using our algorithm.
- Dynamic action recognition /
- continuity and periodicity /
- periodic action /
- functional data analysis (FDA) /
- wearable motion capture system

HTML全文

图 1 窗口提取方法比较

Fig. 1 Comparison of methods about window extraction

下载: 全尺寸图片幻灯片

图 2 4阶傅里叶级数拟合

Fig. 2 Fitting by 4 term Fourier series

下载: 全尺寸图片幻灯片

图 3 不同行为的数据差异

Fig. 3 Differences of data from different actions

下载: 全尺寸图片幻灯片

图 4 导函数图像

Fig. 4 Image of derivative function

下载: 全尺寸图片幻灯片

图 5 左脚数据与腰部数据对比

Fig. 5 Comparison of data from left foot and waist

下载: 全尺寸图片幻灯片

图 6 单测量单元的周期提取结果

Fig. 6 Result of period extraction from data about single measuring unit

下载: 全尺寸图片幻灯片

图 7 周期提取结果

Fig. 7 The results of the extraction of the periodic data

下载: 全尺寸图片幻灯片

图 8 基于WARD的本文方法与传统方法分类结果汇总

Fig. 8 Results of our method and traditional methods based on WARD

下载: 全尺寸图片幻灯片

图 9 动作捕捉系统及传感器位置

Fig. 9 Motion capture system and position of sensors

下载: 全尺寸图片幻灯片

图 10 单个样本数据序列

Fig. 10 A sample of the data series

下载: 全尺寸图片幻灯片

图 11 基于自采集数据库的本文算法与其他方法比较

Fig. 11 Comparison our method with others based on our database

下载: 全尺寸图片幻灯片

表 1 WARD中的行为描述

Table 1 Description of the behaviors in WARD

编号	行为类别	行为描述
1	正常行走	向前走持续超过10秒
2	逆时针行走	逆时针走持续超过10秒
3	顺时针行走	顺时针走持续超过10秒
4	向左转	原地左转持续超过10秒
5	向右转	原地右转持续超过10秒
6	上楼梯	上超过10阶的楼梯
7	下楼梯	下超过10阶的楼梯
8	慢跑	慢跑持续超过10秒
9	跳	原地跳超过5次
10	推轮椅	推轮椅超过10秒

下载: 导出CSV

表 2 10种动态行为类别的混淆矩阵

Table 2 Confusion matrix of 10 dynamic action classes

	1	2	3	4	5	6	7	8	9	10	识别率（%)
1	97	0	0	0	0	1	0	0	0	2	97
2	0	97	0	1	0	0	0	0	0	2	97
3	0	0	98	0	1	0	1	0	0	0	98
4	0	0	0	100	0	0	0	0	0	0	100
5	0	0	0	0	100	0	0	0	0	0	100
6	0	0	0	0	0	97	1	1	1	0	97
7	0	0	0	0	0	2	94	1	0	3	94
8	0	0	0	0	0	0	0	100	0	0	100
9	0	0	0	0	0	1	5	0	94	0	94
10	1	0	0	0	0	0	0	0	1	98	98
总识别率											97.5

下载: 导出CSV

表 3 基于WARD的算法结果对比

Table 3 Comparison with other methods based on WARD

	窗口大小	特征提取方法	分类器	分类平均耗时（s)	识别率(%)
文献[6]	200	时频域特征	DT	< 1	93.7
文献[11]	45	局部保持投影（LPP)	DSC	N/A	87.05
文献[12]	40	随机投影（RP)	SRC	N/A	82.1
本文算法	45	标定起始点的周期提取	SVM	< 1	97.5

下载: 导出CSV

表 4 不同分类器分类结果对比

Table 4 Comparison with other classifiers

分类器	识别率（%)
支持向量机	97.5
决策树	89.5
朴素贝叶斯	86.9
K-近邻	93.9

下载: 导出CSV

参考文献(20)

[1]	Poppe R. A survey on vision-based human action recognition. Image and Vision Computing, 2010, 28(6): 976-990 doi: 10.1016/j.imavis.2009.11.014
[2]	王蒙, 戴亚平, 王庆林.单目视觉下目标三维行为的时间尺度不变建模及识别.自动化学报, 2013, 40(8): 1644-1653 http://www.aas.net.cn/CN/abstract/abstract18432.shtml Wang Meng, Dai Ya-Ping, Wang Qing-Lin. Time-scale invariant modeling and classifying for object behaviors in 3D space based on monocular vision. Acta Automatica Sinica, 2013, 40(8): 1644-1653 http://www.aas.net.cn/CN/abstract/abstract18432.shtml
[3]	Yang J, Wang S Q, Chen N J, Chen X, Shi P F. Wearable accelerometer based extendable activity recognition system. In: Proceedings of the 2010 IEEE International Conference on Robotics and Automation. Anchorage, AK, USA: IEEE, 2010. 3641-3647
[4]	Preece S J, Goulermas J Y, Kenney L P J, Howard D. A comparison of feature extraction methods for the classification of dynamic activities from accelerometer data. IEEE Transactions on Biomedical Engineering, 2009, 56(3): 871-879 doi: 10.1109/TBME.2008.2006190
[5]	He W H, Guo Y C, Gao C, Li X K. Recognition of human activities with wearable sensors. EURASIP Journal on Advances in Signal Processing, 2012, 2012(1): 108 doi: 10.1186/1687-6180-2012-108
[6]	李锋, 潘敬奎.基于三轴加速度传感器的人体运动识别.计算机研究与发展, 2016, 53(3): 621-631 doi: 10.7544/issn1000-1239.2016.20148159 Li Feng, Pan Jing-Kui. Human motion recognition based on Triaxial accelerometer. Journal of Computer Research and Development, 2016, 53(3): 621-631 doi: 10.7544/issn1000-1239.2016.20148159
[7]	Najafi B, Aminian K, Paraschiv-Ionescu A, Loew F, Bula C J, Robert P. Ambulatory system for human motion analysis using a kinematic sensor: monitoring of daily physical activity in the elderly. IEEE Transactions on Biomedical Engineering, 2003, 50(6): 711-723 doi: 10.1109/TBME.2003.812189
[8]	Zhang M, Sawchuk A A. A feature selection-based framework for human activity recognition using wearable multimodal sensors. In: Proceedings of the 6th International Conference on Body Area Networks. Beijing, China: ICST, 2011. 92-98
[9]	Ghasemzadeh H, Jafari R. Physical movement monitoring using body sensor networks: a phonological approach to construct spatial decision trees. IEEE Transactions on Industrial Informatics, 2011, 7(1): 66-77 doi: 10.1109/TII.2010.2089990
[10]	Wang Z L, Jiang M, Hu Y H, Li H Y. An incremental learning method based on probabilistic neural networks and adjustable fuzzy clustering for human activity recognition by using wearable sensors. IEEE Transactions on Information Technology in Biomedicine, 2012, 16(4): 691-699 doi: 10.1109/TITB.2012.2196440
[11]	Yang A Y, Jafari R, Sastry S S, Bajcsy R. Distributed recognition of human actions using wearable motion sensor networks. Journal of Ambient Intelligence and Smart Environments, 2009, 1(2): 103-115
[12]	肖玲, 李仁发, 罗娟.体域网中一种基于压缩感知的人体动作识别方法.电子与信息学报, 2013, 35(1): 119-125 http://www.cnki.com.cn/Article/CJFDTOTAL-DZYX201301018.htm Xiao Ling, Li Ren-Fa, Luo Juan. Recognition of human activity based on compressed sensing in body sensor networks. Journal of Electronics and Information Technology, 2013, 35(1): 119-125 http://www.cnki.com.cn/Article/CJFDTOTAL-DZYX201301018.htm
[13]	Ramsay J O, Silverman B W. Functional Data Analysis. New York, USA: Springer-Verlag, 2005.
[14]	Ramsay J O, Silverman B W. Applied Functional Data Analysis: Methods and Case Studies. New York, USA: Springer, 2002.
[15]	苏为华, 孙利荣, 崔峰.一种基于函数型数据的综合评价方法研究.统计研究, 2013, 30(2): 88-94 http://cdmd.cnki.com.cn/Article/CDMD-10353-1012336072.htm Su Wei-Hua, Sun Li-Rong, Cui Feng. One comprehensive evaluation method study based on functional data. Statistical Research, 2013, 30(2): 88-94 http://cdmd.cnki.com.cn/Article/CDMD-10353-1012336072.htm
[16]	李淳芃, 王兆其, 夏时洪.人体运动的函数数据分析与合成.软件学报, 2009, 20(6): 1664-1672 http://www.cnki.com.cn/Article/CJFDTOTAL-RJXB200906025.htm Li Chun-Peng, Wang Zhao-Qi, Xia Shi-Hong. Motion synthesis for virtual human using functional data analysis. Journal of Software, 2009, 20(6): 1664-1672 http://www.cnki.com.cn/Article/CJFDTOTAL-RJXB200906025.htm
[17]	Chang C C, Lin C J. LIBSVM: a library for support vector machines. ACM Transactions on Intelligent Systems and Technology, 2011, 2(3): Article No. 27
[18]	Boser B B, Guyon I M, Vapnik V N. A training algorithm for optimal margin classifiers. In: Proceedings of the 5th Annual Workshop on Computational Learning Theory. Pittsburgh, Pennsylvania, USA: ACM, 1992. 144-152
[19]	MATLAB中文论坛. MATLAB神经网络30个案例分析.中国北京:北京航空航天大学出版社, 2010. 120-132 http://cdmd.cnki.com.cn/Article/CDMD-11912-2006063682.htm MATLAB Chinese Forum. MATLAB Neural Network 30 Case Analysis. Beijing, China: Beijing University of Aeronautics and Astronautics Press, 2010. 120-132 http://cdmd.cnki.com.cn/Article/CDMD-11912-2006063682.htm
[20]	Wright J, Yang A Y, Ganesh A, Sastry S S, Ma Y. Robust face recognition via sparse representation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, 31(2): 210-227 doi: 10.1109/TPAMI.2008.79