基于CSI小样本学习的场景鲁棒性跌倒检测系统

doi:10.11959/j.issn.2096-3750.2023.00339

物联网学报 ›› 2023, Vol. 7 ›› Issue (2): 118-132.doi: 10.11959/j.issn.2096-3750.2023.00339

基于CSI小样本学习的场景鲁棒性跌倒检测系统

曾钰婷¹, 毕宿志¹^,², 郑莉莉¹, 林晓辉¹, 王晖³

¹ 深圳大学电子与信息工程学院，广东深圳 518060
² 鹏城实验室宽带通信研究部，广东深圳 518066
³ 深圳信息职业技术学院，广东深圳 518172

修回日期:2023-03-16 出版日期:2023-06-30 发布日期:2023-06-01
作者简介:曾钰婷（1997- ），女，深圳大学电子与信息工程学院硕士生，主要研究方向为无线感知
毕宿志（1987- ），男，博士，深圳大学电子与信息工程学院副教授，主要研究方向为无线通信网络资源管理、移动计算与无线感知
郑莉莉（1990- ），女，博士，深圳大学电子与信息工程学院博士后，主要研究方向为无线感知
林晓辉（1975- ），男，博士，深圳大学电子与信息工程学院教授，主要研究方向为无人机通信网络的优化设计
王晖（1969- ），男，博士，深圳信息职业技术学院教授，主要研究方向为无线通信与信号处理
基金资助:
国家自然科学基金资助项目(62271325);鹏城实验室宽带通信研究部重点研究计划项目;广东省教育厅科技重点专项(2020ZDZX3050);广东省基础与应用基础研究基金资助项目(2022A1515011219);广东省基础与应用基础研究基金资助项目(2022A1515010973);深圳市科创委基础研究项目(20220810142637001);深圳市科创委基础研究项目(JCYJ20210324093011030);深圳市科创委基础研究项目(JCYJ20190808120415286);智慧城市物联网国家重点实验室（澳门大学）开放课题项目(SKL-IoTSC(UM)-2021-2023/ORPF/A03/2022)

Scenario-adaptive wireless fall detection system based on few-shot learning

Yuting ZENG¹, Suzhi BI¹^,², Lili ZHENG¹, Xiaohui LIN¹, Hui WANG³

¹ College of Electronics and Information Engineering, Shenzhen University, Shenzhen 518060, China
² Pengcheng Laboratory Broadband Communication Research Department, Shenzhen 518066, China
³ Shenzhen Institute of Information Technology, Shenzhen 518172, China

Revised:2023-03-16 Online:2023-06-30 Published:2023-06-01
Supported by:
The National Natural Science Foundation of China(62271325);The Major Key Project of PCL Department of Broadband Communication;The Key Project of Department of Education of Guangdong Province(2020ZDZX3050);The Guangdong Basic and Applied Basic Research Foundation(2022A1515011219);The Guangdong Basic and Applied Basic Research Foundation(2022A1515010973);The Shenzhen Science and Technology Program(20220810142637001);The Shenzhen Science and Technology Program(JCYJ20210324093011030);The Shenzhen Science and Technology Program(JCYJ20190808120415286);The Open Research Project Programme of the State Key Laboratory of Internet of Things for Smart City (University of Macau)(SKL-IoTSC(UM)-2021-2023/ORPF/A03/2022)

摘要/Abstract

摘要：

采用小样本学习技术设计了基于CSI的场景鲁棒性跌倒检测系统（FDFL, fall detection system based on few-shot learning）。现有基于Wi-Fi无线信道状态信息（CSI, channel state information）的跌倒检测方法跨场景应用性能退化明显，通常需要在每个应用场景采集并标记大量的CSI样本，给大规模部署造成极高的成本。为此，引入了小样本学习的方法，可以在陌生场景标注样本数量不足的情况下仍然保持高准确率的跌倒检测性能。所提FDFL 主要分为源域的元训练和目标域的元学习两个阶段。源域的元训练阶段包含数据预处理和分类训练两个部分，数据预处理阶段将采集的原始CSI幅度和相位数据进行去噪、分段等操作；分类训练阶段利用大量处理好的源域数据样本训练一个基于卷积神经网络的CSI特征提取器。在目标域的元学习阶段，基于元训练模块中训练的特征提取器对目标域中采样的少量标注样本进行有效的特征提取，进而训练生成一个轻量型机器学习分类器对跨场景下的跌倒行为进行检测。通过多个不同场景下的实验，FDFL在只需要目标域少量样本下即可以实现对跌倒、坐着、步行、坐下的四分类任务达到95.52%的平均识别准确率，并且对测试环境、人员目标、设备位置等因素的变化保持鲁棒的检测准确性。

关键词: 无线感知, 跌倒检测, CSI, 跨域检测, 小样本学习

Abstract:

A scenario robust fall detection system based on few-shot learning (FDFL) in wireless environment was designed.The performance of existing fall detection methods based on Wi-Fi channel state information (CSI) degrades significantly across scenarios, which requires collecting and marking a large number of CSI samples in each application scenario, resulting in high cost for large-scale deployment.Therefore, the method of few-shot learning was introduced, which can maintain the performance of fall detection with high accuracy when the number of annotated samples in unfa-miliar scenes is insufficient.The proposed FDFL was mainly divided into two stages, source domain meta-training and target domain meta-learning.The meta training stage of the source domain consists of two parts: data preprocessing and classification training.In the data preprocessing stage, the collected original CSI amplitude and phase data were denoised and segmented.In the classification training stage, a large number of processed source domain data samples were used to train a CSI feature extractor based on convolutional neural network.In the meta-learning stage of the target domain, the limited labeled data sampled in the target domain was effectively extracted based on the feature extractor trained in the meta-training module, and then a lightweight machine learning classifier was trained to detect the fall behavior under the cross-scene.Through several experiments in different scenarios, FDFL can achieve an average accuracy of 95.52% for the four classification tasks of falling, sitting, walking and sit down with only a small number of samples in the target domain, and maintain robust detection accuracy for changes in test environment, personnel target and equipment location.

Key words: Wi-Fi sensing, fall detection, CSI, cross-domain detection, few-shot learning

中图分类号:

TP391.7

曾钰婷, 毕宿志, 郑莉莉, 林晓辉, 王晖. 基于CSI小样本学习的场景鲁棒性跌倒检测系统[J]. 物联网学报, 2023, 7(2): 118-132.

Yuting ZENG, Suzhi BI, Lili ZHENG, Xiaohui LIN, Hui WANG. Scenario-adaptive wireless fall detection system based on few-shot learning[J]. Chinese Journal on Internet of Things, 2023, 7(2): 118-132.

图/表 12

图1

图2

图3

图4

图5

图6

表1

表2

表3

表4

图7

图8

参考文献 33

[1]	预防老年人跌倒健康教育核心信息[J]. 江苏卫生保健, 2022(2): 50-52.
	Prevention of falls in the elderly health education core information[J]. Jiangsu Health Care, 2022(2): 50-52.
[2]	HU Y Q , ZHANG F , WU C S ,et al. DeFall:environment-independent passive fall detection using Wi-Fi[J]. IEEE Internet of Things Journal, 2022,9(11): 8515-8530.
[3]	WANG Y X , WU K S , NI L M . WiFall:device-free fall detection by wireless networks[J]. IEEE Transactions on Mobile Computing, 2017,16(2): 581-594.
[4]	ZHANG D Q , WANG H , WANG Y S ,et al. Anti-fall:A non-intrusive and real-time fall detector leveraging CSI from commodity Wi-Fi devices[C]// Proceedings of International Conference on Smart Homes and Health Telematics. Cham:Springer, 2015: 181-193.
[5]	WANG H , ZHANG D Q , WANG Y S ,et al. RT-fall:a real-time and contactless fall detection system with commodity Wi-Fi devices[J]. IEEE Transactions on Mobile Computing, 2017,16(2): 511-526.
[6]	PALIPANA S , ROJAS D , AGRAWAL P ,et al. FallDeFi[J]. Proceedings of the ACM on Interactive,Mobile,Wearable and Ubiquitous Technologies, 2018,1(4): 1-25.
[7]	THAMMASAT E , CHAICHARN J . A simply fall-detection algorithm using accelerometers on a smartphone[C]// Proceedings of 5th 2012 Biomedical Engineering International Conference. Piscataway:IEEE Press, 2013: 1-4.
[8]	KHALILI A , SOLIMANAH , ASADUZZAMANM ,et al. Wi-Fi sensing:applications and challenges[J]. The Journal of Engineering, 2020,2020(3): 87-97.
[9]	PATWARI N , WILSON J , ANANTHANARAYANAN S ,et al. Monitoring breathing via signal strength in wireless networks[J]. IEEE Transactions on Mobile Computing, 2014,13(8): 1774-1786.
[10]	XIANG P , JI P , ZHANG D . Enhance RSS-based indoor localization accuracy by leveraging environmental physical features[J]. Wireless Communications and Mobile Computing, 2018(1).
[11]	ZHENG X L , WANG J L , SHANGGUANL F ,et al. Design and implementation of a CSI-based ubiquitous smoking detection system[J]. IEEE/ACM Transactions on Networking, 2017,25(6): 3781-3793.
[12]	KONINGS D , GRACE R , ALAM F . A stacked neural network-based machine learning framework to detect activities and falls within multiple indoor environments using Wi-Fi CSI[J]. IEEE Sensors Letters, 2021,5(5): 1-4.
[13]	LIN Z Z , XIE Y C , GUO X N ,et al. WiEat:fine-grained device-free eating monitoring leveraging Wi-Fi signals[C]// Proceedings of 2020 29th International Conference on Computer Communications and Networks (ICCCN). Piscataway:IEEE Press, 2020: 1-9.
[14]	ZHANG J , WU F X , HU W ,et al. WiEnhance:towards data augmentation in human activity recognition using Wi-Fi signal[C]// Proceedings of 2019 15th International Conference on Mobile Ad-Hoc and Sensor Networks (MSN). Piscataway:IEEE Press, 2020: 309-314.
[15]	HE Y , CHEN Y , HU Y ,et al. Wi-Fi vision:sensing,recognition,and detection with commodity MIMO-OFDM Wi-Fi[J]. IEEE Internet of Things Journal, 2020,7(9): 8296-8317.
[16]	CHENG X Y , HUANG B K , ZONG J . A device-free human fall detection system based on GMM-HMM using Wi-Fi signals[C]// Proceedings of 2021 IEEE 4th International Conference on Electronics and Communication Engineering (ICECE). Piscataway:IEEE Press, 2022: 87-92.
[17]	KEENANR M , TRAN L N . Fall detection using Wi-Fi signals and threshold-based activity segmentation[C]// Proceedings of 2020 IEEE 31st Annual International Symposium on Personal,Indoor and Mobile Radio Communications. Piscataway:IEEE Press, 2020: 1-6.
[18]	WANG Y C , YANG S , LI F ,et al. FallViewer:a fine-grained indoor fall detection system with ubiquitous Wi-Fi devices[J]. IEEE Internet of Things Journal, 2021,8(15): 12455-12466.
[19]	WU M J , TANG Z L . Research on user action recognition method based on parallel CNN-BiLSTM neural network[C]// Proceedings of 2021 IEEE 4th Advanced Information Management,Communicates,Electronic and Automation Control Conference (IMCEC). Piscataway:IEEE Press, 2021: 2003-2009.
[20]	NAKAMURA T , BOUAZIZI M , YAMAMOTO K ,et al. Wi-Fi-CSIbased fall detection by spectrogram analysis with CNN[C]// Proceedings of GLOBECOM 2020 - 2020 IEEE Global Communications Conference. Piscataway:IEEE Press, 2021: 1-6.
[21]	CHOI H , FUJIMOTO M , MATSUI T ,et al. Wi-CaL:Wi-Fi sensing and machine learning based device-free crowd counting and localization[J]. IEEE Access, 2022(10): 24395-24410.
[22]	LIU Z X , YUANR H , YUANY Z ,et al. A sensor-free crowd counting frame work for indoor environments based on channel state information[J]. IEEE Sensors Journal, 2022,22(6): 6062-6071.
[23]	TANG Z L , LIU Q Q , WU M J ,et al. Wi-Fi CS Igesture recognitionbased on parallel LSTM-FCN deep space-time neural network[J]. China Communications, 2021,18(3): 205-215.
[24]	SCOTTT R , RIDGEWAY K , MOZERM C . Adapted deep embeddings:asynthesis of methods for k-shot inductive transfer learning[C]// Proceedings of the 32nd International Conference on Neural Information Processing Systems. New York:ACM, 2018: 76-85.
[25]	ZHANG Y , CHEN Y , WANG Y J ,et al. CSI-based human activity recognition with graph few-shot learning[J]. IEEE Internet of Things Journal, 2022,9(6): 4139-4151.
[26]	MA X R , ZHAO Y N , ZHANG L ,et al. Practical device-free gesture recognition using Wi-Fi signals based on meta learning[J]. IEEE Transactions on Industrial Informatics, 2020,16(1): 228-237.
[27]	SUN Q R , LIU Y Y , CHUA T S ,et al. Meta-transfer learning for few-shot learning[C]// Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2020: 403-412.
[28]	SHI Z G , ZHANG J A , XU R Y ,et al. Environment-robust device-free human activity recognition with channel-state-information enhancement and one-shot learning[J]. IEEE Transactions on Mobile Computing, 2022,21(2): 540-554.
[29]	DING X , JIANG T , ZHONG Y ,et al. Improving WiFi-based human activity recognition with adaptive initial state via one-shot learning[C]// Proceedings of 2021 IEEE Wireless Communications and Networking Conference (WCNC). Piscataway:IEEE Press, 2021: 1-6.
[30]	WANG Y J , YAO L , WANG Y ,et al. Robust CSI-based human activity recognition with augment few shot learning[J]. IEEE Sensors Journal, 2021,21(21): 24297-24308.
[31]	WANG D Z , YANG J F , CUIW ,et al. CAUTION:a robust WiFi-based human authentication system via few-shot open-set recognition[J]. IEEE Internet of Things Journal, 2022,9(18): 17323-17333.
[32]	SADREAZAMI H , BOLIC M , RAJAN S . TL-FALL:contactless indoor fall detection using transfer learning from a pretrained model[C]// Proceedings of 2019 IEEE International Symposium on Medical Measurements and Applications (MeMeA). Piscataway:IEEE Press, 2019: 1-5.
[33]	FDIDA S , PAU G , KASERA S ,et al. Proceedings of the 21st annual ACM International Conference on Mobile Computing and Networking (Mobicom)[EB]. 2015.

		源域场景A→目标域场景A	场景A→场景B	场景A→场景C	场景A→场景D	平均值
1-shot	二分类	0.791 0	0.801 0	0.923 1	0.867 6	0.845 7
	三分类	0.720 2	0.701 2	0.818 6	0.780 9	0.755 2
	四分类	0.661 4	0.636 3	0.658 7	0.837 9	0.698 6
5-shot	二分类	0.859 3	0.867 8	0.979 7	0.993 9	0.925 2
	三分类	0.864 4	0.858 8	0.901 1	0.892 5	0.879 2
	四分类	0.850 8	0.828 8	0.905 2	0.884 8	0.867 4
15-shot	二分类	0.947 4	0.943 5	1.000 0	1.000 0	0.972 7
	三分类	0.935 0	0.941 7	0.983 6	0.942 4	0.950 7
	四分类	0.922 8	0.927 2	0.961 9	0.978 4	0.947 6

		源域场景A→目标域场景A	场景A→场景B	场景A→场景C	场景A→场景D	平均值
1-shot	二分类	0.607 0	0.632 1	0.755 9	0.858 2	0.713 3
	三分类	0.600 9	0.635 5	0.662 1	0.747 3	0.661 5
	四分类	0.557 7	0.586 1	0.618 4	0.669 2	0.607 9
5-shot	二分类	0.862 7	0.906 8	0.994 9	0.947 5	0.928 0
	三分类	0.852 0	0.861 0	0.896 6	0.900 5	0.877 5
	四分类	0.838 1	0.854 2	0.905 2	0.900 4	0.874 5
15-shot	二分类	0.963 9	0.943 5	1.000 0	0.999 4	0.976 7
	三分类	0.958 0	0.947 5	0.983 1	0.972 7	0.965 3
	四分类	0.955 3	0.926 3	0.957 1	0.982 0	0.955 2

		源域场景A→目标域场景A	场景A→场景B	场景A→场景C	场景A→场景D	平均值
1-shot	二分类	0.791 0	0.801 0	0.923 1	0.867 6	0.845 7
	三分类	0.705 7	0.674 5	0.820 9	0.773 0	0.743 5
	四分类	0.666 4	0.632 1	0.670 7	0.825 5	0.698 7
5-shot	二分类	0.883 1	0.910 2	0.925 4	0.959 8	0.919 6
	三分类	0.866 7	0.822 6	0.849 4	0.907 0	0.861 4
	四分类	0.831 4	0.847 5	0.852 2	0.895 5	0.856 7
15-shot	二分类	0.956 8	0.929 9	0.996 5	0.987 6	0.967 7
	三分类	0.941 8	0.909 0	0.975 8	0.969 2	0.949 0
	四分类	0.946 5	0.912 3	0.952 4	0.935 7	0.936 7

基于CSI小样本学习的场景鲁棒性跌倒检测系统

Scenario-adaptive wireless fall detection system based on few-shot learning

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 33

相关文章 5

Metrics

推荐阅读 0

房间	场景A				场景B				场景C				场景D
动作类型	跌倒	坐着	步行	坐下	跌倒	坐着	步行	坐下	跌倒	坐着	步行	坐下	跌倒	坐着	步行	坐下
实验次数	300	300	300	300	120	120	120	120	120	120	120	120	120	120	120	120

[1]	马柱华, 罗丽平. 非理想顺序干扰消除和信道状态信息下SWIPT-NOMA-CR网络中断性能[J]. 物联网学报, 2023, 7(1): 129-139.
[2]	魏忠诚, 张新秋, 连彬, 王巍, 赵继军. 基于Wi-Fi信号的身份识别技术研究[J]. 物联网学报, 2021, 5(4): 107-119.
[3]	单少伟,杨卫东,肖乐,王珂. Wi-Pest：一种基于CSI的储粮害虫检测方法[J]. 物联网学报, 2020, 4(4): 51-61.
[4]	陈慕涵,郭佳佳,李潇,金石. 基于深度学习的大规模MIMO信道状态信息反馈[J]. 物联网学报, 2020, 4(1): 33-44.
[5]	廖勇,姚海梅,花远肖. 一种基于深度学习的物联网信道状态信息获取算法[J]. 物联网学报, 2019, 3(1): 8-13.