基于小波特征匹配的短时人体行为识别

doi:10.16182/j.issn1004731x.joss.22-0176

摘要/Abstract

摘要：

在人体行为识别研究中，特征选取是关键。为了获取充分且稳定的行为特征，往往对超过一个行为周期的长时行为数据进行处理，而小于一个行为周期的短时行为数据特征通常不稳定，难以做到准确稳定识别。提出一种基于小波变换和模板匹配相结合的短时人体行为识别方法。使用小波变换方法提取系数特征，将短时测试样本的特征与模板库中的特征进行匹配，根据相似性对行为动作做出分类识别。实验结果表明，此方法对于短时行为动作具有较为准确和稳定的识别性能，有助于实现对人体行为动作的实时识别。

关键词: 短时行为, 小波变换, 近似系数, 模板匹配, 1范数

Abstract:

The selection of features is the key problem in the study of human activity recognition. In order to obtain sufficient and stable behavioral features, long-time behavioral data that exceed one behavior cycle are often processed, while short-time behavioral data with less than one behavioral cycle are usually unstable, making it difficult to achieve accurate and stable identification. This paper proposes a short-time human activity recognition method based on the combination of wavelet transform and template matching. Coefficient features are extracted using wavelet transform method. The features of the short-time test samples are matched with the features in the template library to make classification recognition of activity based on similarity. The experimental results show that this method has more accurate and stable recognition performance for short-time behavioral activity, which helps to realize real-time recognition of human behavioral activity.

Key words: short-time activity, wavelet transform, approximate coefficient, template matching, 1 norm

中图分类号:

TP391.4

苏本跃, 张利, 何清旋, 盛敏. 基于小波特征匹配的短时人体行为识别[J]. 系统仿真学报, 2023, 35(1): 158-168.

Benyue Su, Li Zhang, Qingxuan He, Min Sheng. Short-time Human Activity Recognition Based on Wavelet Features Matching[J]. Journal of System Simulation, 2023, 35(1): 158-168.

图/表 15

图1

图2

图3

图4

表1

表2

图5

图6

图7

表3

用户独立实验结果

动作时长/s	特征	分类方法	平均识别率/%
动作时长/s	特征	分类方法	WARD1.0数据集	自采集数据集
大约0.13	最大值、最小值、均值、方差	K-近邻	49.30	30.12
		决策树	33.15	45.00
		支持向量机	40.85	37.88
		朴素贝叶斯	25.35	25.63
		随机森林	59.25	52.00
	小波变换的近似系数 $w$	模板匹配	78.65	71.00

表3

图8

图9

表4

用户依赖实验结果

动作时长/s	特征	分类方法	平均识别率/%
动作时长/s	特征	分类方法	WARD1.0数据集	自采集数据集
大约0.13	最大值、最小值、均值、方差	K-近邻	51.50	47.97
		决策树	33.55	49.22
		支持向量机	33.85	40.63
		朴素贝叶斯	27.35	28.13
		随机森林	53.70	55.16
	小波变换的近似系数 $w$	模板匹配	90.45	96.41

表4

表5

表6

参考文献 28

1	Yi W J, Sarkar O, Mathavan S, et al. Wearable Sensor Data Fusion for Remote Health Assessment and Fall Detection[C]// IEEE International Conference on Electro/Information Technology. Milwaukee, WI, USA: IEEE, 2014: 303-307.
2	Zhang H, Guo Y, Zanotto D. Accurate Ambulatory Gait Analysis in Walking and Running Using Machine Learning Models[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering (S1534-4320), 2019, 28(1): 191-202.
3	Hsu Y L, Yang S C, Chang H C, et al. Human Daily and Sport Activity Recognition Using a Wearable Inertial Sensor Network[J]. IEEE Access (S2169-3536), 2018, 6: 31715-31728.
4	Faridee A Z M, Ramamurthy S R, Hossain H M S, et al. Happyfeet: Recognizing and Assessing Dance on the Floor[C]// the 19th International Workshop on Mobile Computing Systems & Applications. New York, United States, 2018: 49-54.
5	Dang L M, Min K, Wang H, et al. Sensor-Based and Vision-based Human Activity Recognition: A Comprehensive Survey[J]. Pattern Recognition (S0031-3203), 2020, 108: 107561.
6	郭毅博, 孟文化, 范一鸣, 等. 基于可穿戴传感器数据的人体行为识别数据特征提取方法[J]. 计算机辅助设计与图形学学报, 2021, 33(8): 1246-1253.
	Guo Yibo, Meng Wenhua, Fan Yiming, et al. Wearable Sensor Data Based Human Behavior Recognition: A Method of Data Feature Extraction[J]. Journal of Computer-Aided Design & Computer Graphics, 2021, 33(8): 1246-1253.
7	Zhong R, Rau P L P, Yan X. Application of Smart Bracelet to Monitor Frailty-Related Gait Parameters of Older Chinese Adults: A Preliminary Study[J]. Geriatrics & Gerontology International (S1444-1586), 2018, 18(9): 1366-1371.
8	盛敏, 刘双庆, 王婕, 等. 基于改进模板匹配的智能下肢假肢运动意图实时识别[J]. 控制与决策, 2020, 35(9): 2153-2161.
	Sheng Min, Liu Shuangqing, Wang Jie, et al. Real-Time Motion Intent Recognition of Intelligent Lower Limb Prosthesis Based on Improved Template Matching Technique[J]. Control and Decision, 2020, 35(9): 2153-2161.
9	Rosati S, Balestra G, Knaflitz M. Comparison of Different Sets of Features for Human Activity Recognition by Wearable Sensors[J]. Sensors (S1424-8220), 2018, 18(12): 4189.
10	Ma Z, Yang L T, Lin M, et al. Weighted Support Tensor Machines for Human Activity Recognition with Smartphone Sensors[J]. IEEE Transactions on Industrial Informatics (S1551-3203), 2021: 1-9.
11	Jain A, Kanhangad V. Human Activity Classification in Smartphones Using Accelerometer and Gyroscope Sensors[J]. IEEE Sensors Journal (S1530-437X), 2017, 18(3): 1169-1177.
12	Tao D, Jin L, Yuan Y, et al. Ensemble Manifold Rank Preserving for Acceleration-Based Human Activity Recognition[J]. IEEE Transactions on Neural Networks and Learning Systems (S2162-237X), 2014, 27(6): 1392-1404.
13	Zheng X, Wang M, Ordieres-Meré J. Comparison of Data Preprocessing Approaches for Applying Deep Learning to Human Activity Recognition in the Context of Industry 4.0[J]. Sensors (S1424-8220), 2018, 18(7): 2146.
14	Wang F, Gong W, Liu J. On Spatial Diversity in Wifi-based Human Activity Recognition: A Deep Learning based Approach[J]. IEEE Internet of Things Journal (S2327-4662), 2018, 6(2): 2035-2047.
15	苏本跃, 郑丹丹, 汤庆丰, 等. 单传感器数据驱动的人体日常短时行为识别方法[J]. 红外与激光工程, 2019, 48(2): 282-290.
	Su Benyue, Zheng Dandan, Tang Qingfeng, et al. Human Daily Short-Time Activity Recognition Method Driven by Single Sensor Data[J]. Infrared and Laser Engineering, 2019, 48(2): 282-290.
16	唐向宏, 李齐良. 时频分析与小波变换[M]. 北京: 科学出版社, 2016.
	Tang Xianghong, Li Qiliang. Time-Frequency Analysis and Wavelet Transform[M]. Beijing: Science Press, 2016.
17	Abdu-Aguye M G, Gomaa W. Competitive Feature Extraction for Activity Recognition Based on Wavelet Transforms and Adaptive Pooling[C]// 2019 International Joint Conference on Neural Networks (IJCNN). Budapest, Hungary: IEEE, 2019: 1-8.
18	Wang N, Ambikairajah E, Lovell N H, et al. Accelerometry Based Classification of Walking Patterns Using Time-Frequency Analysis[C]// 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. Lyon, France: IEEE, 2007: 4899-4902.
19	He H, Tan Y, Zhang W. A Wavelet Tensor Fuzzy Clustering Scheme for Multi-sensor Human Activity Recognition[J]. Engineering Applications of Artificial Intelligence (S0952-1976), 2018, 70: 109-122.
20	Yang H, Huang C, Wang F, et al. Large-Scale and Rotation-Invariant Template Matching Using Adaptive Radial Ring Code Histograms[J]. Pattern Recognition (S0031-3203), 2019, 91: 345-356.
21	Li L J, Su H, Lim Y, et al. Object Bank: An Object-Level Image Representation for High-level Visual Recognition[J]. International Journal of Computer Vision (S0920-5691), 2014, 107(1): 20-39.
22	Sadanand S, Corso J J. Action Bank: A High-Level Representation of Activity in Video[C]// 2012 IEEE Conference on Computer Vision and Pattern Recognition. Providence, RI, USA: IEEE, 2012: 1234-1241.
23	Yang A Y, Jafari R, Sastry S S, et al. Distributed Recognition of Human Actions Using Wearable Motion Sensor Networks[J]. Journal of Ambient Intelligence and Smart Environments (S1876-1364), 2009, 1(2): 103-115.
24	朱红蕾, 朱昶胜, 徐志刚. 人体行为识别数据集研究进展[J]. 自动化学报, 2018, 44(6): 978-1004.
	Zhu Honglei, Zhu Changsheng, Xu Zhigang. Research Advances on Human Activity Recognition Datasets[J]. Acta Automatica Sinica, 2018, 44(6): 978-1004.
25	Ignatov A. Real-time Human Activity Recognition from Accelerometer Data Using Convolutional Neural Networks[J]. Applied Soft Computing (S1568-4946), 2018, 62: 915-922.
26	Ortega-Anderez D, Lotfi A, Langensiepen C, et al. A Multi-Level Refinement Approach Towards the Classification of Quotidian Activities Using Accelerometer Data[J]. Journal of Ambient Intelligence and Humanized Computing (S1868-5137), 2019, 10(11): 4319-4330.
27	Lu J, Zheng X, Sheng M, et al. Efficient Human Activity Recognition Using a Single Wearable Sensor[J]. IEEE Internet of Things Journal (S2327-4662), 2020, 7(11): 11137-11146.
28	Su B, Tang Q, Jiang J, et al. A Novel Method for Short-Time Human Activity Recognition Based on Improved Template Matching Technique[C]// the 15th ACM SIGGRAPH Conference on Virtual-Reality Continuum and Its Applications in Industry, Volume 1, 2016: 233-242.

行为类别	行为描述	WARD1.0数据集	自采集数据集
上楼	上超过10阶楼梯	√	√
下楼	下超过10阶楼梯	√	√
正常行走	向前走10 s以上	√	√
顺时针转弯	顺时针转弯10 s以上	√	√
逆时针转弯	逆时针转弯10 s以上	√	√
左转弯	左转弯10 s以上	√	√
右转弯	右转弯10 s以上	√	√
慢跑	慢跑10 s以上	√	√
跳跃	原地跳超过5次	√
推轮椅	推轮椅超过10 s	√

参数	WARD1.0数据集	自采集数据集	PAMAP2数据集
动作种类	10	8	10
频率/Hz	30	96	100
传感器数量	5	5	3
特征维度	5	6	6
样本数量	1 000	320	3 500

文献	特征	方法分类	数据集	动作时长/s	识别率/%
[9]	均值, 方差, 熵谱等	支持向量机	文献[9] 数据集	5	97.10
[25]	最大值, 最小值, 均值等	卷积神经网络	WISDM	1	90.42
[25]	最大值, 最小值, 均值等	卷积神经网络	UCI	1	94.35
[26]	均值, 熵, 能量等	随机森林	文献[26]数据集	1	99.28
[27]	最大值, 最小值, 均值等	梯度增强决策树	DaLiAc	5	93.00
[10]	均值, 方差, 相关系数等	加权支持张量机	UCI	>10	89.40
本文方法	小波变换的近似系数	模板匹配	WARD1.0	0.13	90.45
			自采集数据	0.13	96.41
			PAMAP2	0.13	90.03

动作时长/s	方法	数据集	匹配时间/s	识别率/%
大约0.13	文献[28]	WARD1.0	6.94	95.15
	文献[28]	自采集数据	2.99	96.10
	本文方法	WARD1.0	1.95	90.45
	本文方法	自采集数据	0.62	96.41

[1]	陈涛, 王艳, 纪志成. 基于经验小波变换的光伏功率组合预测模型[J]. 系统仿真学报, 2021, 33(11): 2627-2635.
[2]	黄升, 张涛, 黄崇君, 李玉梅, 邓虎, 张霞. 井下数据获取及粘滑特征分析[J]. 系统仿真学报, 2019, 31(11): 2517-2526.
[3]	王依人, 邓国庆, 夏营威, 张龙, 刘勇, 张文. 基于方向可调滤波器的血管图像增强算法[J]. 系统仿真学报, 2018, 30(6): 2095-2101.
[4]	闫国强, 周宁宁, 张少白. 基于MMTD与小波硬阈值的脑电信号去噪方法[J]. 系统仿真学报, 2018, 30(4): 1490-1495.
[5]	张鸿喜, 张洁. 小波变换在舷外有源诱饵仿真模型验证中的应用[J]. 系统仿真学报, 2018, 30(1): 318-324.
[6]	赵迪, 徐志胜. 基于多分辨率奇异熵的混凝土缺陷检测[J]. 系统仿真学报, 2017, 29(4): 761-766.
[7]	周国雄, 周先雁, 王解军, 黄特. 基于小波分析—蚁群BP网络的木构件缺陷无损检测[J]. 系统仿真学报, 2015, 27(11): 2804-2810.
[8]	李豪, 李锐华, 胡波, 潘玲, 郭其一. 基于导波的大电机定子绝缘损伤定位仿真研究[J]. 系统仿真学报, 2015, 27(11): 2816-2821.