论文部分内容阅读
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。
Aimed at the problem of strong non-stationary, multi-scale and low signal-to-noise ratio (SNR) of acoustic signals radiated by vehicles moving in dynamic noisy environment, a vehicle model sounding based on Local Mean Decomposition (LMD) and local projection energy Feature extraction method. Firstly, LMD method is used to adaptively decompose the collected acoustic signals to obtain the product of the PFs at each scale, and to separate the PF components containing the characteristic frequency components from the strong background noise. Secondly, the LMD decomposition results are processed Weighted optimization, reconstruct the feature PF component, filter out the false component and the weakly related component to enhance the feature information. Finally, the energy of the feature PF component is projected equidistantly into the energy gathering area, and the low-dimensional features of the acoustic signal are constructed based on the energy scale Vector, and recognize the eigenvector through the learning of artificial neural network. On the test dataset, the LMD local projection energy characteristics are used to identify the vehicle, and different intensities of noise are added to the experimental dataset. LMD decomposition and local projection energy calculation are performed. The calculation results are compared with those obtained by other feature extraction methods Comparative analysis. The results show that this method is very sensitive to vehicle model information, and the recognition rate is 93.4%. It can effectively suppress the background noise interference in dynamic environment and obtain target-sensitive narrow-band signal with good noise immunity. Projection in the energy accumulation area can effectively remove redundant features and improve the real-time performance of the algorithm.