VEHICLE ENGINE SOUND ANALYSIS APPLIED TO TRAFFIC CONGESTION ESTIMATION
VEHICLE ENGINE SOUND ANALYSIS APPLIED TO TRAFFIC CONGESTION ESTIMATION
复制标题
汽车发动机声音分析应用于交通拥堵估计
DOI：
--
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
P. Rao
中科院分区：
文献类型：
--
作者：
Nikhil Bhave;P. Rao
据报道，对与道路车辆交通有关的声学特征进行了调查。与车辆类型和运动状态相关的可计算特征可用于监控交通拥堵。在目前的工作中，研究了不同车辆（大致分为两轮、三轮车和重型车辆）的声学特征。发动机声音的源滤波器模型用于导出合适的特征。通过 k-NN 分类器在手动标记的交通声音数据库上将基于共振峰的特征的性能与梅尔频率倒谱系数 (MFCC) 的性能进行比较。虽然鸣喇叭的存在可以用来指示交通状况，但由于某些区域可能限制鸣喇叭，因此它并不可靠。我们考虑车辆的其他非喇叭声学，以寻找自动检测交通状态的新颖解决方案。道路上行驶的车辆发出的声音具有独特的运动状态和车辆类型，通常很容易识别。因此，从交通噪声中提取的声学特征有可能为交通状况提供重要线索。根据人类声音感知的知识，此类声学线索预计在车辆声音的短时幅度频谱中突出。在过去的车辆识别工作中已经探索了谱域表示 (2)。然而，车辆声音的频谱受到车身结构及其运动状态（静态、移动、加速等）的影响。基于小波的方法（3）使用六阶样条小波变换，通过对照记录和处理的声学信号的现有数据库分析声学特征，在存在其他噪声时检测任意类型的车辆的到达。为了确保尽可能少的错误匹配，使用小波包系数块之间的能量分布以及随机搜索接近最佳足迹的过程来构建训练数据库。在基于频谱的方法中，(4)提取与车辆发动机相关的基频，并建立基频、发动机气缸数及其RPM之间的关系。它确定了基频和谐波谱结构在车辆分类中的作用。 (5) 中描述的使用反向传播的车辆分类算法是另一种基于谱域的算法
An investigation of acoustic features relating to vehicular traffic on roadways is reported. Computable features that relate to the type of vehicle and state of motion can be useful in monitoring traffic congestion. In the present work, different vehicles, broadly classified into two, three wheelers and heavy vehicle, are studied for their acoustic signatures. A source filter model of engine sound is used to derive suitable features. The performance of formant based features is compared with that of Mel-frequency cepstral coefficients (MFCC) via a k-NN classifier on a manually labelled database of traffic sounds. While the presence of honking can be used to indicate the condition of traffic, it is not reliable due to the possibility of restrictions on honking in certain zones. We consider other non-honk acoustics of the vehicle to look for novel solutions to automatic detection of the state of the traffic. Vehicles moving on the road create sounds that are distinctive of the state of motion as well as vehicle type, often easily identifiable. Acoustic features extracted from traffic noise thus have the potential of providing important clues to traffic conditions. From the knowledge of human sound perception, such acoustic cues are expected to be prominent in the short-time magnitude frequency spectrum of the vehicle sound. Spectral domain representations have been explored in past work on vehicle recognition (2). The spectrum of the vehicle sound however is influenced by the structure of the vehicle body as well as its state of motion (static, moving, accelerating, etc.). A wavelet based method (3), uses a sixth order spline wavelet transform to detect arrival of vehicles of arbitrary type when other noises are present by analysis of acoustic signatures against an existing database of recorded and processed acoustic signals. To ensure least possible false matching, a training database is constructed using the distribution of energies among blocks of wavelet packet coefficients with a procedure for random search for a near-optimal footprint. Among spectral based methods, (4) extracts the fundamental frequency associated with the engine of the vehicle, and establishes a relation between the fundamental frequency, number of cylinders of the engines and their RPM. It identifies the role of this fundamental frequency and the harmonic spectral structure in classification of vehicles. Vehicle classification algorithm using back propagation described in (5) is another spectral domain based