基于LMD局部投影能量特征的车型识别
|
摘要
针对动态噪声环境下行进中的机动车辐射出的声信号具有强非平稳性、多尺度性及低信噪比的问题,提出一种基于局部均值分解(LMD)和局部投影能量计算的车型声特征提取方法。首先,利用LMD方法对采集的声信号进行自适应分解,得到各尺度上的乘积函数(PF)分量,从强背景噪声中分离出包含车型特征频率成分的PF分量;其次,对LMD分解结果进行加权优化,重构特征PF分量,滤除虚假成分及弱相关分量,以增强特征信息;最后,将特征PF分量的能量等距离投影到能量聚集区内,基于能量尺度构造声信号的低维特征向量,并通过人工神经网络的学习对特征向量进行识别。在试验数据集上,采用LMD局部投影能量特征对目标车辆进行车型识别,并对试验数据集添加不同强度的噪声,进行LMD分解及局部投影能量计算,将计算结果与其他特征提取方法计算结果进行对比分析。结果表明:该方法对于车型信息十分敏感,识别率达到93.4%;可以有效抑制动态环境下的背景噪声干扰,获取目标敏感的窄带信号,具有很好的抗噪能力;选择在重构窄带信号的能量聚集区内进行投影计算,可以有效去除冗余特征,同时提高算法的实时性。
Under dynamic noisy environment,acoustic signals radiated from moving vehicles exhibit characteristics of non-stationary,multi-scale and low signal to noise ratio.Aiming at this problem,a feature extraction approach based on local mean decomposition(LMD)and local projective energy calculation was proposed.Firstly,LMD was utilized to decompose the collected acoustic signal adaptively into different PF components at various scales.In this way the PFs that contained characteristic frequency components related to vehicle types can be separated from intense background noise.Secondly,a weight optimization method was performed on the chosen PF components to reconstruct a new characteristic PF component in an attempt to reinforce informative features.In this way,the false and unrelated components could be removed.At last,energy of the reconstructed PF was equidistantly projected to the special frequency band,where most energy was accumulated.Then the corresponding energy distribution in this area was usedto construct the feature vector of vehicle acoustic signals,whose dimension was reduced effectively.Artificial neural networks(ANN)were utilized to identify vehicle types based on the acquired feature vectors.On experiment dataset,local LMD projection energy feature was adopted to identify vehicle type,and noises in different degree were added to dataset to decompose by LMD and calculate local projection energy.The calculated results were compared with results of different feature extraction methods.The results show that the proposed method can suppress the interruption of background noises in the dynamic environment as much as possible,obtain the narrowband signals that are sensitive to targets and show well anti-noise capacity.The projection calculation of the local energy of the obtained signals at the energy focus region will get rid of some redundant characteristics,and ensure the real-time performance of the algorithm.
Under dynamic noisy environment,acoustic signals radiated from moving vehicles exhibit characteristics of non-stationary,multi-scale and low signal to noise ratio.Aiming at this problem,a feature extraction approach based on local mean decomposition(LMD)and local projective energy calculation was proposed.Firstly,LMD was utilized to decompose the collected acoustic signal adaptively into different PF components at various scales.In this way the PFs that contained characteristic frequency components related to vehicle types can be separated from intense background noise.Secondly,a weight optimization method was performed on the chosen PF components to reconstruct a new characteristic PF component in an attempt to reinforce informative features.In this way,the false and unrelated components could be removed.At last,energy of the reconstructed PF was equidistantly projected to the special frequency band,where most energy was accumulated.Then the corresponding energy distribution in this area was usedto construct the feature vector of vehicle acoustic signals,whose dimension was reduced effectively.Artificial neural networks(ANN)were utilized to identify vehicle types based on the acquired feature vectors.On experiment dataset,local LMD projection energy feature was adopted to identify vehicle type,and noises in different degree were added to dataset to decompose by LMD and calculate local projection energy.The calculated results were compared with results of different feature extraction methods.The results show that the proposed method can suppress the interruption of background noises in the dynamic environment as much as possible,obtain the narrowband signals that are sensitive to targets and show well anti-noise capacity.The projection calculation of the local energy of the obtained signals at the energy focus region will get rid of some redundant characteristics,and ensure the real-time performance of the algorithm.
引文
[1]COIFMAN B,KIM S.Speed Estimation and Length Based Vehicle Classification from Freeway Singleloop Detectors[J].Transportation Research Part C,2009,17(4):349-364.
[2]田寅,董宏辉,贾利民,等.用于车辆分类的多传感器车型特征融合算法[J].华南理工大学学报:自然科学版,2014,42(3):52-58.TIAN Yin,DONG Hong-hui,JIA Li-min,et al.Multisensor Signature Fusion Algorithm for Vehicle Type Classification[J].Journal of South China University of Technology:Natural Science Edition,2014,42(3):52-58.
[3]KAFAI M,BHANU B.Dynamic Bayesian Networks for Vehicle Classification in Video[J].IEEE Transactions on Industrial Informatics,2012,8(1):100-109.
[4]YANG B,LEI Y.Vehicle Detection and Classification for Low-speed Congested Traffic with Anisotropic Magnetoresistive Sensor[J].IEEE Sensors Journal,2015,15(2):1132-1138.
[5]方向,丁凯,齐世福,等.地面装甲目标声信号的混沌特征提取[J].解放军理工大学学报:自然科学版,2013,14(2):218-221.FANG Xiang,DING Kai,QI Shi-fu,et al.Chaotic Feature Extraction of Acoustic Signals from Armored Vehicles[J].Journal of PLA University of Science and Technology:Natural Science Edition,2013,14(2):218-221.
[6]WU H,MENDEL J M.Classification of Battlefield Ground Vehicles Using Acoustic Features and Fuzzy Logic Rule-based Classifiers[J].IEEE Transactions on Fuzzy Systems,2007,15(1):56-72.
[7]李开明,张群,罗迎,等.地面车辆目标识别研究综述[J].电子学报,2014,42(3):538-546.LI Kai-ming,ZHANG Qun,LUO Ying,et al.Review of Ground Vehicles Recognition[J].ACTA Electronica Sinica,2014,42(3):538-546.
[8]刘立君,周滨涛,戴鸿滨,等.管道MIG焊双通道电弧声信号采集与特征分析[J].焊接学报,2012,33(8):41-44.LIU Li-jun,ZHOU Bin-tao,DAI Hong-bin,et al.Dual-channel Signal Acquisition and Characteristics Analysis of Arc in Pipe MIG Welding[J].Transactions of the China Welding Institution,2012,33(8):41-44.
[9]陈果,于明月,刘永泉,等.基于倒频谱分析的航空发动机转静碰摩部位识别[J].机械工程学报,2014,50(7):32-38.CHEN Guo,YU Ming-yue,LIU Yong-quan,et al.Identifying Rotor-stator Rubbing Positions Using the Cepstrum Analysis Technique[J].Journal of Mechanical Engineering,2014,50(7):32-38.
[10]蒋志迪.一种基于小波分解和压缩感知的冲击声学无损检测方法[J].电子与信息学报,2012,34(12):3021-3026.JIANG Zhi-di.A New Impact-acoustics Non-destructive Test Method Based on Wavelet Decomposition and Compressive Sensing[J].Journal of Electronics&Information Technology,2012,34(12):3021-3026.
[11]汤健,柴天佑,丛秋梅,等.基于EMD和选择性集成学习算法的磨机负荷参数软测量[J].自动化学报,2014,40(9):1853-1865.TANG Jian,CHAI Tian-you,CONG Qiu-mei,et al.Soft Sensor Approach for Modeling Mill Load Parameters Based on EMD and Selective Ensemble Learning Algorithm[J].ACTA Automation Sinica,2014,40(9):1853-1865.
[12]罗向龙,高静怀,牛国宏,等.交通事件小波分解与支持向量机声频识别方法[J].交通运输工程学报,2010,10(2):116-121.LUO Xiang-long,GAO Jing-huai,NIU Guo-hong,et al.Traffic Incident Acoustic Recognition Method Based on Wavelet Decomposition and Support Vector Machine[J].Journal of Traffic and Transportation Engineering,2010,10(2):116-121.
[13]李京华,赵易峰,许家栋.基于子波能量和神经网络分类器的机动车车型识别[J].中国公路学报,2007,20(3):97-102.LI Jing-hua,ZHAO Yi-feng,XU Jia-dong.Vehicle Type Recognition Based on Wavelet Energy and Neural Network Classifier[J].China Journal of Highway and Transport,2007,20(3):97-102.
[14]齐晓轩,纪建伟,韩晓微.机动车声信号能量特征的小波包提取[J].沈阳农业大学学报,2011,42(3):372-375.QI Xiao-xuan,JI Jian-wei,HAN Xiao-wei.Energy Feature Extraction of Vehicle Acoustic Signal by Wavelet Packet[J].Journal of Shenyang Agricultural University,2011,42(3):372-375.
[15]罗向龙,牛国宏,吴潜蛟,等.基于经验模态分解和支持向量机的车型声频识别[J].应用声学,2010,29(3):178-183.LUO Xiang-long,NIU Guo-hong,WU Qian-jiao,et al.Vehicle Recognition by Acoustic Signals Based on EMD and SVM[J].Applied Acoustic,2010,29(3):178-183.
[16]SMITH J S.The Local Mean Decomposition Method and Its Application to EEG Perception Data[J].Journal of the Royal Society Interface,2005,2(5):443-454.
[17]JIANG W,ZHENG Z,ZHU Y,et al.Demodulation for Hydraulic Pump Fault Signals Based on Local Mean Decomposition and Improved Adaptive Multiscale Morphology Analysis[J].Mechanical Systems and Signal Processing,2015,58-59:179-205.
[18]WANG Y,MARKERT R,XIANG J,et al.Research on Variational Mode Decomposition and Its Application in Detecting Rub-impact Fault of the Rotor System[J].Mechanical Systems and Signal Processing,2015,60-61:243-251.
[19]OBUCHOWSKI J,WYLOMANSKA A,ZIMROZ R.Selection of Informative Frequency Band in Local Damage Detection in Rotating Machinery[J].Mechanical Systems and Signal Processing,2014,48(1/2):138-152.
[20]LI Y,XU M,WANG R,et al.A fault Diagnosis Scheme for Rolling Bearing Based on Local Mean Decomposition and Improved Multiscale Fuzzy Entropy[J].Journal of Sound and Vibration,2016,360:277-299.
[21]雷旭,侯莹莹,武奇生.地磁车辆检测器车型分类算法设计[J].长安大学学报:自然科学版,2013,33(5):118-123.LEI Xu,HOU Ying-ying,WU Qi-sheng.Design of Classification Algorithm on Vehicle Geomagnetic Detecting System[J].Journal of Chang’an University:Natural Science Edition,2013,33(5):118-123.
[2]田寅,董宏辉,贾利民,等.用于车辆分类的多传感器车型特征融合算法[J].华南理工大学学报:自然科学版,2014,42(3):52-58.TIAN Yin,DONG Hong-hui,JIA Li-min,et al.Multisensor Signature Fusion Algorithm for Vehicle Type Classification[J].Journal of South China University of Technology:Natural Science Edition,2014,42(3):52-58.
[3]KAFAI M,BHANU B.Dynamic Bayesian Networks for Vehicle Classification in Video[J].IEEE Transactions on Industrial Informatics,2012,8(1):100-109.
[4]YANG B,LEI Y.Vehicle Detection and Classification for Low-speed Congested Traffic with Anisotropic Magnetoresistive Sensor[J].IEEE Sensors Journal,2015,15(2):1132-1138.
[5]方向,丁凯,齐世福,等.地面装甲目标声信号的混沌特征提取[J].解放军理工大学学报:自然科学版,2013,14(2):218-221.FANG Xiang,DING Kai,QI Shi-fu,et al.Chaotic Feature Extraction of Acoustic Signals from Armored Vehicles[J].Journal of PLA University of Science and Technology:Natural Science Edition,2013,14(2):218-221.
[6]WU H,MENDEL J M.Classification of Battlefield Ground Vehicles Using Acoustic Features and Fuzzy Logic Rule-based Classifiers[J].IEEE Transactions on Fuzzy Systems,2007,15(1):56-72.
[7]李开明,张群,罗迎,等.地面车辆目标识别研究综述[J].电子学报,2014,42(3):538-546.LI Kai-ming,ZHANG Qun,LUO Ying,et al.Review of Ground Vehicles Recognition[J].ACTA Electronica Sinica,2014,42(3):538-546.
[8]刘立君,周滨涛,戴鸿滨,等.管道MIG焊双通道电弧声信号采集与特征分析[J].焊接学报,2012,33(8):41-44.LIU Li-jun,ZHOU Bin-tao,DAI Hong-bin,et al.Dual-channel Signal Acquisition and Characteristics Analysis of Arc in Pipe MIG Welding[J].Transactions of the China Welding Institution,2012,33(8):41-44.
[9]陈果,于明月,刘永泉,等.基于倒频谱分析的航空发动机转静碰摩部位识别[J].机械工程学报,2014,50(7):32-38.CHEN Guo,YU Ming-yue,LIU Yong-quan,et al.Identifying Rotor-stator Rubbing Positions Using the Cepstrum Analysis Technique[J].Journal of Mechanical Engineering,2014,50(7):32-38.
[10]蒋志迪.一种基于小波分解和压缩感知的冲击声学无损检测方法[J].电子与信息学报,2012,34(12):3021-3026.JIANG Zhi-di.A New Impact-acoustics Non-destructive Test Method Based on Wavelet Decomposition and Compressive Sensing[J].Journal of Electronics&Information Technology,2012,34(12):3021-3026.
[11]汤健,柴天佑,丛秋梅,等.基于EMD和选择性集成学习算法的磨机负荷参数软测量[J].自动化学报,2014,40(9):1853-1865.TANG Jian,CHAI Tian-you,CONG Qiu-mei,et al.Soft Sensor Approach for Modeling Mill Load Parameters Based on EMD and Selective Ensemble Learning Algorithm[J].ACTA Automation Sinica,2014,40(9):1853-1865.
[12]罗向龙,高静怀,牛国宏,等.交通事件小波分解与支持向量机声频识别方法[J].交通运输工程学报,2010,10(2):116-121.LUO Xiang-long,GAO Jing-huai,NIU Guo-hong,et al.Traffic Incident Acoustic Recognition Method Based on Wavelet Decomposition and Support Vector Machine[J].Journal of Traffic and Transportation Engineering,2010,10(2):116-121.
[13]李京华,赵易峰,许家栋.基于子波能量和神经网络分类器的机动车车型识别[J].中国公路学报,2007,20(3):97-102.LI Jing-hua,ZHAO Yi-feng,XU Jia-dong.Vehicle Type Recognition Based on Wavelet Energy and Neural Network Classifier[J].China Journal of Highway and Transport,2007,20(3):97-102.
[14]齐晓轩,纪建伟,韩晓微.机动车声信号能量特征的小波包提取[J].沈阳农业大学学报,2011,42(3):372-375.QI Xiao-xuan,JI Jian-wei,HAN Xiao-wei.Energy Feature Extraction of Vehicle Acoustic Signal by Wavelet Packet[J].Journal of Shenyang Agricultural University,2011,42(3):372-375.
[15]罗向龙,牛国宏,吴潜蛟,等.基于经验模态分解和支持向量机的车型声频识别[J].应用声学,2010,29(3):178-183.LUO Xiang-long,NIU Guo-hong,WU Qian-jiao,et al.Vehicle Recognition by Acoustic Signals Based on EMD and SVM[J].Applied Acoustic,2010,29(3):178-183.
[16]SMITH J S.The Local Mean Decomposition Method and Its Application to EEG Perception Data[J].Journal of the Royal Society Interface,2005,2(5):443-454.
[17]JIANG W,ZHENG Z,ZHU Y,et al.Demodulation for Hydraulic Pump Fault Signals Based on Local Mean Decomposition and Improved Adaptive Multiscale Morphology Analysis[J].Mechanical Systems and Signal Processing,2015,58-59:179-205.
[18]WANG Y,MARKERT R,XIANG J,et al.Research on Variational Mode Decomposition and Its Application in Detecting Rub-impact Fault of the Rotor System[J].Mechanical Systems and Signal Processing,2015,60-61:243-251.
[19]OBUCHOWSKI J,WYLOMANSKA A,ZIMROZ R.Selection of Informative Frequency Band in Local Damage Detection in Rotating Machinery[J].Mechanical Systems and Signal Processing,2014,48(1/2):138-152.
[20]LI Y,XU M,WANG R,et al.A fault Diagnosis Scheme for Rolling Bearing Based on Local Mean Decomposition and Improved Multiscale Fuzzy Entropy[J].Journal of Sound and Vibration,2016,360:277-299.
[21]雷旭,侯莹莹,武奇生.地磁车辆检测器车型分类算法设计[J].长安大学学报:自然科学版,2013,33(5):118-123.LEI Xu,HOU Ying-ying,WU Qi-sheng.Design of Classification Algorithm on Vehicle Geomagnetic Detecting System[J].Journal of Chang’an University:Natural Science Edition,2013,33(5):118-123.