非接触式身份识别的深度学习算法
|
摘要
针对传统基于Wi-Fi的身份识别方法手工编码特征效率低、准确率不高的问题,提出一种基于深度学习的非接触式身份识别(WiID)算法。该算法通过分析子载波中信道状态信息数据的空间相关性,建立了用于深度学习的输入矩阵;采用二维卷积运算从相邻子载波中提取局部空间特征;构建门限循环单元层,从时间维度对空间特征进行时序建模,完成空间与时间两个维度的步态特征提取,实现端到端的非接触式身份识别,有效减少了数据预处理工作量。实验结果表明,与卷积神经网络和循环神经网络算法相比,该算法识别准确率得到了有效提高;在6种不同的实验场景下,该算法的身份识别准确率介于92.9%～95.6%之间,具有良好的身份识别效果及算法鲁棒性。
A contactless human identification algorithm WiID based on deep learning is proposed to solve the problem that traditional Wi-Fi-based human identification methods have low efficiency of manual feature extraction and low accuracy. An input matrix for deep learning is established by analyzing the spatial correlation of channel state information data in subcarriers. The two-dimensional convolution operation is used to extract local spatial features from adjacent subcarriers, and spatial features are modeled from the temporal dimension through the gated recurrent unit layer. Gait feature extractions in both spatial and temporal dimensions are performed, and the end-to-end contactless human identification is realized to effectively reduce the workload of data preprocessing. Experimental results and comparisons with the convolutional neural network and recurrent neural network show that the identification accuracy of the proposed algorithm is effectively improved. The identification accuracy of the proposed algorithm under six different scenarios ranges from 92.9% to 95.6%, and the algorithm has good identification effect and robustness.
A contactless human identification algorithm WiID based on deep learning is proposed to solve the problem that traditional Wi-Fi-based human identification methods have low efficiency of manual feature extraction and low accuracy. An input matrix for deep learning is established by analyzing the spatial correlation of channel state information data in subcarriers. The two-dimensional convolution operation is used to extract local spatial features from adjacent subcarriers, and spatial features are modeled from the temporal dimension through the gated recurrent unit layer. Gait feature extractions in both spatial and temporal dimensions are performed, and the end-to-end contactless human identification is realized to effectively reduce the workload of data preprocessing. Experimental results and comparisons with the convolutional neural network and recurrent neural network show that the identification accuracy of the proposed algorithm is effectively improved. The identification accuracy of the proposed algorithm under six different scenarios ranges from 92.9% to 95.6%, and the algorithm has good identification effect and robustness.
引文
[1]UNAR J A,SENG W C,ABBASI A.A review of biometric technology along with trends and prospects[J].Pattern Recognition,2014,47(8):2673-2688.
[2]CONNOR P,ROSS A.Biometric recognition by gait:a survey of modalities and features[J].Computer Vision and Image Understanding,2018,167:1-27.
[3]WU Zifeng,HUANG Yongzhen,WANG Liang,et al.A comprehensive study on cross-view gait based human identification with deep CNNs[J].IEEETransactions on Pattern Analysis&Machine Intelligence,2016,39(2):209-226.
[4]ZHANG Yuting,PAN Gang,JIA Kui,et al.Accelerometer-based gait recognition by sparse representation of signature points with clusters[J].IEEE Transactions on Cybernetics,2015,45(9):1864-1875.
[5]AL-NAIMI I,WONG C B,MOORE P,et al.Multimodal approach for non-tagged indoor identification and tracking using smart floor and pyroelectric infrared sensors[J].International Journal of Computational Science and Engineering,2017,14(1):1-15.
[6]XIN Tong,GUO Bin,WANG Zhu,et al.Freesense:indoor human identification with Wi-Fi signals[C]∥Proceedings of the IEEE Global Communications Conference.Piscataway,NJ,USA:IEEE,2016:1-7.
[7]ZENG Y,PATHAK P H,MOHAPATRA P.WiWho:WiFi-based person identification in smart spaces[C]∥Proceedings of the 15th International Conference on Information Processing in Sensor Networks.Piscataway,NJ,USA:IEEE,2016:1-12.
[8]ZHANG Jin,WEI Bo,HU Wen,et al.Wifiid:human identification using WiFi signal[C]∥Proceedings of the International Conference on Distributed Computing in Sensor Systems.Piscataway,NJ,USA:IEEE,2016:75-82.
[9]OHARA K,MAEKAWA T,MATSUSHITA Y.Detecting state changes of indoor everyday objects using Wi-Fi channel state information[J].Proceedings of the ACM on Interactive,Mobile,Wearable and Ubiquitous Technologies,2017,1(3):88.
[10]ALI K,LIU A X,WANG W,et al.Keystroke recognition using WiFi signals[C]∥Proceedings of the 21st Annual International Conference on Mobile Computing and Networking.New York,USA:ACM,2015:90-102.
[11]CHO K,MERRIENBOER B,GULCEHRE C,et al.Learning phrase representations using RNN encoderde-coder for statistical machine translation[EB/OL].(2014-09-03)[2018-05-16].http:∥arxiv.org/abs/1406.1078.
[12]杨铮,刘云浩.Wi-Fi雷达:从RSSI到CSI[J].中国计算机学会通讯,2014,10(11):55-60.YANG Zheng,LIU Yunhao.Wi-Fi radar:from RSSIto CSI[J].Communications of the CCF,2014,10(11):55-60.
[13]GOODFELLOW I,BENGIO Y,COURVILLE A,et al.Deep learning[M].Cambridge,UK:MIT Press,2016:322-323.
[14]ZHANG Kai,ZUO Wangmeng,CHEN Yunjin,et al.Beyond a Gaussian denoiser:residual learning of deep CNN for image denoising[J].IEEE Transactions on Image Processing,2017,26(7):3142-3155.
[15]IOFFE S,SZEGEDY C.Batch normalization:accelerating deep network training by reducing internal covariate shift[C]∥Proceedings of the 32nd International Conference on Machine Learning.New York,USA:ICML,2015:448-456.
[16]JOZEFOWICZ R,ZAREMBA W,SUTSKEVER I.An empirical exploration of recurrent network architectures[C]∥Proceedings of the 32nd International Conference on Machine Learning.New York,USA:IC-ML,2015:2342-2350.
[17]ZAREMBA W,SUTSKEVER I,VINYALS O.Recurrent neural network regularization[EB/OL].(2015-02-19)[2018-06-05].http:∥arxiv.org/abs/1409.2329.
[18]滕千礼,ESMAEILI KELISHOMI A,蔡忠闽.采用运动传感器的人体运动识别深度模型[J].西安交通大学学报,2018,52(8):60-66.TENG Qianli,ESMAEILI KELISHOMI A,CAIZhongmin.A deep learning model for human activity recognition using motion sensor data[J].Journal of Xi’an Jiaotong University.2018,52(8):60-66.
[19]TOLIAS G,SICRE R,JEGOU H.Particular object retrieval with integral max-pooling of CNN activations[EB/OL].(2016-02-24)[2018-06-05].http:∥arxiv.org/abs/1511.05879.
[20]HALPERIN D,HU W,SHETH A,et al.Tool release:gathering 802.11ntraces with channel state information[J].ACM SIGCOMM Computer Communication Review,2011,41(1):53-53.
[2]CONNOR P,ROSS A.Biometric recognition by gait:a survey of modalities and features[J].Computer Vision and Image Understanding,2018,167:1-27.
[3]WU Zifeng,HUANG Yongzhen,WANG Liang,et al.A comprehensive study on cross-view gait based human identification with deep CNNs[J].IEEETransactions on Pattern Analysis&Machine Intelligence,2016,39(2):209-226.
[4]ZHANG Yuting,PAN Gang,JIA Kui,et al.Accelerometer-based gait recognition by sparse representation of signature points with clusters[J].IEEE Transactions on Cybernetics,2015,45(9):1864-1875.
[5]AL-NAIMI I,WONG C B,MOORE P,et al.Multimodal approach for non-tagged indoor identification and tracking using smart floor and pyroelectric infrared sensors[J].International Journal of Computational Science and Engineering,2017,14(1):1-15.
[6]XIN Tong,GUO Bin,WANG Zhu,et al.Freesense:indoor human identification with Wi-Fi signals[C]∥Proceedings of the IEEE Global Communications Conference.Piscataway,NJ,USA:IEEE,2016:1-7.
[7]ZENG Y,PATHAK P H,MOHAPATRA P.WiWho:WiFi-based person identification in smart spaces[C]∥Proceedings of the 15th International Conference on Information Processing in Sensor Networks.Piscataway,NJ,USA:IEEE,2016:1-12.
[8]ZHANG Jin,WEI Bo,HU Wen,et al.Wifiid:human identification using WiFi signal[C]∥Proceedings of the International Conference on Distributed Computing in Sensor Systems.Piscataway,NJ,USA:IEEE,2016:75-82.
[9]OHARA K,MAEKAWA T,MATSUSHITA Y.Detecting state changes of indoor everyday objects using Wi-Fi channel state information[J].Proceedings of the ACM on Interactive,Mobile,Wearable and Ubiquitous Technologies,2017,1(3):88.
[10]ALI K,LIU A X,WANG W,et al.Keystroke recognition using WiFi signals[C]∥Proceedings of the 21st Annual International Conference on Mobile Computing and Networking.New York,USA:ACM,2015:90-102.
[11]CHO K,MERRIENBOER B,GULCEHRE C,et al.Learning phrase representations using RNN encoderde-coder for statistical machine translation[EB/OL].(2014-09-03)[2018-05-16].http:∥arxiv.org/abs/1406.1078.
[12]杨铮,刘云浩.Wi-Fi雷达:从RSSI到CSI[J].中国计算机学会通讯,2014,10(11):55-60.YANG Zheng,LIU Yunhao.Wi-Fi radar:from RSSIto CSI[J].Communications of the CCF,2014,10(11):55-60.
[13]GOODFELLOW I,BENGIO Y,COURVILLE A,et al.Deep learning[M].Cambridge,UK:MIT Press,2016:322-323.
[14]ZHANG Kai,ZUO Wangmeng,CHEN Yunjin,et al.Beyond a Gaussian denoiser:residual learning of deep CNN for image denoising[J].IEEE Transactions on Image Processing,2017,26(7):3142-3155.
[15]IOFFE S,SZEGEDY C.Batch normalization:accelerating deep network training by reducing internal covariate shift[C]∥Proceedings of the 32nd International Conference on Machine Learning.New York,USA:ICML,2015:448-456.
[16]JOZEFOWICZ R,ZAREMBA W,SUTSKEVER I.An empirical exploration of recurrent network architectures[C]∥Proceedings of the 32nd International Conference on Machine Learning.New York,USA:IC-ML,2015:2342-2350.
[17]ZAREMBA W,SUTSKEVER I,VINYALS O.Recurrent neural network regularization[EB/OL].(2015-02-19)[2018-06-05].http:∥arxiv.org/abs/1409.2329.
[18]滕千礼,ESMAEILI KELISHOMI A,蔡忠闽.采用运动传感器的人体运动识别深度模型[J].西安交通大学学报,2018,52(8):60-66.TENG Qianli,ESMAEILI KELISHOMI A,CAIZhongmin.A deep learning model for human activity recognition using motion sensor data[J].Journal of Xi’an Jiaotong University.2018,52(8):60-66.
[19]TOLIAS G,SICRE R,JEGOU H.Particular object retrieval with integral max-pooling of CNN activations[EB/OL].(2016-02-24)[2018-06-05].http:∥arxiv.org/abs/1511.05879.
[20]HALPERIN D,HU W,SHETH A,et al.Tool release:gathering 802.11ntraces with channel state information[J].ACM SIGCOMM Computer Communication Review,2011,41(1):53-53.