| 645 | 11 | 12 |
| 下载次数 | 被引频次 | 阅读次数 |
目的:解决在使用毫米波雷达点云信息做人体动作识别时，传统基于阈值方法鲁棒性差的问题。方法:首先用基于密度的DBSCAN聚类算法分割目标点云与噪点；然后将点云信息图像化处理，构造点云的速度-时间、距离-时间两类散点图；最后搭建特征融合卷积神经网络，以动作的两类特征作为判别依据，对蹲下、摔倒、坐下、行走四类动作识别。结果:在相同测试集中对两类方法进行对比，本文方法的识别准确率比传统基于阈值的方法高25.1%,且在不同训练集占比情况下，对已知个体和未知个体的预测准确率都大于90.3%。结论:将点云数据图像化处理并结合特征融合卷积神经网络的方法能克服身高差异以及干扰动作的影响。
Abstract：Aims: This paper aims to solve the problem of poor robustness of traditional threshold-based methods when using millimeter wave radar point cloud information for human activity recognition. Methods: Firstly, the density-based DBSCAN clustering algorithm was used to segment the target point cloud and noise. Then, the point cloud information was imaged to construct two kinds of scatterplots: velocity-time and distance-time. Finally, a feature fusion convolution neural network was built to recognize squatting, falling, sitting and walking based on the two types of motion features. Results: The recognition accuracy of this method was 25.1% higher than that of the traditional threshold-based methods. And in the case of different proportions of training sets, the prediction accuracy of known individuals and unknown individuals was greater than 90.3%. Conclusions: The method of image processing of point cloud data, combined with feature fusion convolution neural networks, can overcome the influence of height difference and interference actions.
[1] WANG H,ZHAO J,LI J,et al.Wearable sensor-based human activity recognition using hybrid deep learning techniques[J].Security and communication Networks,2020(1):1-12.
[2] EHATISHAM-UL-HAQ M,JAVED A,AZAM M A,et al.Robust human activity recognition using multimodal feature-level fusion[J].IEEE Access,2019,7:60736-60751.
[3] ANDRADE-AMBRIZ Y A,LEDESMA S,IBARRA-MANZANO M A,et al.Human activity recognition using temporal convolutional neural network architecture[J].Expert Systems with Applications,2022,191:116287(1-7).
[4] 朱云鹏，黄希，黄嘉兴.基于3D CNN的人体动作识别研究[J].现代电子技术，2020,43(18):150-152,156.ZHU Y P,HUANG X,HUANG J X.Human action recognition based on 3D CNN[J].Modern Electronics Technique,2020,43(18):150-152,156.
[5] LI H,HE X,CHEN X,et al.Wimotion:A robust human activity recognition using WiFi signals[J].IEEE Access,2019,7:153287-153299.
[6] ZHANG J,WU F,WEI B,et al.Data augmentation and dense-LSTM for human activity recognition using WiFi signal[J].IEEE Internet of Things Journal,2020,8(6):4628-4641.
[7] LI X,HE Y,JING X.A survey of deep learning-based human activity recognition in radar[J].Remote Sensing,2019,11(9):1068(1-22).
[8] GURBUZ S Z,AMIN M G.Radar-based human-motion recognition with deep learning:Promising applications for indoor monitoring[J].IEEE Signal Processing Magazine,2019,36(4):16-28.
[9] 李俊侠，张秦，郑桂妹.超宽带雷达人体姿态识别综述[J].计算机工程与应用，2021,57(3):14-23.LI J X,ZHANG T,ZHENG G M.Overview of human posture recognition by ultra-wideband radar[J].Computer Engineering and Applications,2021,57(3):14-23.
[10] 张丽丽，刘博，屈乐乐，等.基于特征融合卷积神经网络的FMCW雷达人体动作识别[J].电讯技术，2022,62(2):147-154.ZHANG L L,LIU B,QU L Let al.Humanactivity recognition with FMCW radar based on fusion feature convolutional neural network[J].Telecommunication Engineering,2022,62(2):147-154.
[11] 元志安，周笑宇，刘心溥，等.基于RDSNet的毫米波雷达人体跌倒检测方法[J].雷达学报，2021,10(4):656-664.YUAN Z A,ZHOU X Y,LIU X B,et al.Human fall detection method using millimeter-wave radar based on RDSNet[J].Journal of Radars,2021,10(4):656-664.
[12] 谢晓兰，陈梓涵.基于时间距离像的人体动作深度学习分类[J].桂林理工大学学报，2019,39(1):197-203.XIE X L,CHEN Z H.Deep learning for human action classification based on time-range profiles[J].Journal of Guilin University of Technology,2019,39(1):197-203.
[13] ABDULATIF S,AZIZ F,KLEINER B,et al.Real-time capable micro-doppler signature decomposition of walking human limbs[C]//2017 IEEE Radar Conference (RadarConf).Seattle,WA,USA:IEEE,2017:1093-1098.
[14] 丁晨旭，张远辉，孙哲涛，等.基于FMCW雷达的人体复杂动作识别[J].雷达科学与技术，2020,18(6):584-590.DING C X,ZHANG Y H,SUN Z T,etal.human activity classificationbased in FMCW radar[J].Radar Science and Technology,2020,18(6):584-590.
[15] LANG Y,WANG Q,YANG Y,et al.Unsupervised domain adaptation for micro-doppler human motion classification via feature fusion[J].IEEE Geoscience and Remote Sensing Letters,2018,16(3):392-396.
[16] POUR EBRAHIM M,SARVI M,YUCE M R.A doppler radar system for sensing physiological parameters in walking and standing positions[J].Sensors,2017,17(3):485(1-15).
[17] 屈乐乐，张丁元，杨天虹，等.基于双频段FMCW雷达的人体动作识别[J].电讯技术，2022,62(1):59-65.QU L L,ZHANG D Y,YANG T H,et al.Human motion recognition based on dual-band FMCW radar[J].Telecommunication Engineering,2022,62(1):59-65.
[18] PAN H,LIU K,SHEN L,et al.Real-time multi-target activity recognition based on FMCW radar[M]//Intelligent Equipment,Robots,and Vehicles.Singapore:Springer,2021:446-455.
基本信息:
DOI：
中图分类号:TN957.52
引用信息:
[1]田钰琪,刘康,张远辉.基于毫米波雷达点云的人体动作识别[J].中国计量大学学报,2023,34(01):66-73+83.
基金信息:
中德无人驾驶中心毫米波雷达联合实验室基金项目(No.03103-211179); 浙江省自然科学基金项目(No.LY19F010007)