基于时间距离像的人体动作深度学习分类
|
摘要
在人体目标的雷达自动识别中,常用提取到的人体微多普勒特征对人体运动状态或动作姿态进行识别。受人体多变姿态和躯干强回波的影响,人体微多普勒特征有时是微弱和模糊的,难以稳定提取并用于分类。本文使用超宽带雷达录取了人体动作的高分辨率距离像,由连续多帧距离像构建了覆盖整个动作的时间-距离像,采用深度卷积神经网络自动学习时间-距离像的分层特征并进行了分类,对9种动作的平均分类精度达到了96.67%。实验结果验证了深度卷积神经网络对基于时间-距离像的人体动作分类是可行和有效的。
The human micro-Doppler signatures(MDS) are often used to recognize the human kinestate or posture in radar automatic target recognition. Influenced by the body postures or trunk echoes, the MDS are sometimes weak or fuzzy, and difficult to be extracted from radar echoes for human classification. In this paper, an ultra-wideband radar is employed to gather the high resolution range profiles(HRRPs) of human actions, and then form the time-range profiles with a set of the HRRPs from the same action. By automatically learning the hierarchical features of time-range profiles, deep convolutional neural network(DCNN) is adopted to achieve an average accuracy of 96.67% on classification of nine-class human activities. Experimental results demonstrate the feasibility and effectiveness of the DCNN classification for the human action based on time-range profile.
The human micro-Doppler signatures(MDS) are often used to recognize the human kinestate or posture in radar automatic target recognition. Influenced by the body postures or trunk echoes, the MDS are sometimes weak or fuzzy, and difficult to be extracted from radar echoes for human classification. In this paper, an ultra-wideband radar is employed to gather the high resolution range profiles(HRRPs) of human actions, and then form the time-range profiles with a set of the HRRPs from the same action. By automatically learning the hierarchical features of time-range profiles, deep convolutional neural network(DCNN) is adopted to achieve an average accuracy of 96.67% on classification of nine-class human activities. Experimental results demonstrate the feasibility and effectiveness of the DCNN classification for the human action based on time-range profile.
引文
[1]Chen V C,Li F Y,Ho S S,et al.Micro-Doppler effect in radar:phenomenon,model,and simulation study[J].IEEE Transactions on Aerospace and Electronic Systems,2006,42(1):2-21.
[2]庄钊文,刘永祥,黎湘.目标微动特性研究进展[J].电子学报,2007,35(3):520-525.
[3]Kim Y W,Ling H.Human activity classification based on micro-Doppler signatures using a support vector machine[J].IEEE Transactions on Geoscience and Remote Sensing,2009,47(5):1328-1337.
[4]Narayanan R M,Zenaldin M.Radar micro-Doppler signatures of various human activities[J].IET Radar,Sonar & Navigation,2015,9(9):1205-1215.
[5]Erol B,Amin M G,Zhou Z C,et al.Range information for reducing fall false alarms in assisted living[C]//2016 IEEE Radar Conference(Radarconf16).New York:IEEE Express Conference Publishing,2016.DOI:10.11091RADAR.2016.7485176.
[6]Feng B,Chen B,Liu H W.Radar HRRP target recognition with deep networks[J].Pattern Recognition,2017,61:379-393.
[7]Chen S Z,Wang H P,Xu F,et al.Target classification using the deep convolutional networks for SAR images[J].IEEE Transactions on Geoscience and Remote Sensing,2016,54(8):4806-4817.
[8]Kim Y W,Toomajian B.Hand gesture recognition using micro-Doppler signatures with convolutional neural network[J].IEEE Access,2016,4:7125-7130.
[9]Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
[2]庄钊文,刘永祥,黎湘.目标微动特性研究进展[J].电子学报,2007,35(3):520-525.
[3]Kim Y W,Ling H.Human activity classification based on micro-Doppler signatures using a support vector machine[J].IEEE Transactions on Geoscience and Remote Sensing,2009,47(5):1328-1337.
[4]Narayanan R M,Zenaldin M.Radar micro-Doppler signatures of various human activities[J].IET Radar,Sonar & Navigation,2015,9(9):1205-1215.
[5]Erol B,Amin M G,Zhou Z C,et al.Range information for reducing fall false alarms in assisted living[C]//2016 IEEE Radar Conference(Radarconf16).New York:IEEE Express Conference Publishing,2016.DOI:10.11091RADAR.2016.7485176.
[6]Feng B,Chen B,Liu H W.Radar HRRP target recognition with deep networks[J].Pattern Recognition,2017,61:379-393.
[7]Chen S Z,Wang H P,Xu F,et al.Target classification using the deep convolutional networks for SAR images[J].IEEE Transactions on Geoscience and Remote Sensing,2016,54(8):4806-4817.
[8]Kim Y W,Toomajian B.Hand gesture recognition using micro-Doppler signatures with convolutional neural network[J].IEEE Access,2016,4:7125-7130.
[9]Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.