Video visualization for moving sound sources based on binoculars stereo and acoustic holography

The video visualization methods for moving sound sources are very important. The spatial positions in the sound filed measurement of moving sources arc measured by an automated method based on the binoculars stereo technique,the spatial coordinates of the surface of the moving sound source are determined automatically.The sound field of the surface of the moving source is reconstructed with a microphone array using the acoustic holography method for moving sound sources.The spatial mapping between the sound field and the video images is constructed,and also the time sequence of the sound field and the images is founded. Then,the reconstructed sound field and the real-world images are merged,and the visualized sound field video of the moving source is generated automatically.A sound field measurement system based on this method is developed,the measurement results of the moving sound source with this system indicate that the sound field video visualization is realized effectively with the proposed method,the sound sources and their changing process can be seen directly from the generated video,the identification and localization for the moving sound sources can be easier with this method.     

基于声全息方法和双目视觉实现运动声源声场可视化

运动声源声场的可视化是一种重要的运动声源定位的技术手段,利用双目视觉测量技术实现运动声源声场空间的自动测量,自动确定运动声源表面的空间位置,针对声源表面,利用传声器阵列,基于声全息方法实现运动声源声场的重建,建立视频图像与声场的空间映射,并建立视频与声场之间的时序,实现实景视频图像与声场重建结果的融合,可以自动生成声源运动过程的视频。基于该方法所开发了一套试验测量系统,对运动声源的测量试验结果表明,该方法可以有效实现运动声源的视频可视化,使人可以直接从视频中看到声源及其变化过程,使声源的定位和识别变得更加简单。      

Quantitative measurement of pass-by noise radiated by vehicles running at high speeds

It has been a challenge in the past to accurately locate and quantify the pass-by noise source radiated by the running vehicles. A system composed of a microphone array is developed in our current work to do this work. An acoustic-holography method for moving sound sources is designed to handle the Doppler effect effectively in the time domain. The effective sound pressure distribution is reconstructed on the surface of a running vehicle. The method has achieved a high calculation efficiency and is able to quantitatively measure the sound pressure at the sound source and identify the location of the main sound source. The method is also validated by the simulation experiments and the measurement tests with known moving speakers. Finally, the engine noise, tire noise, exhaust noise and wind noise of the vehicle running at different speeds are successfully identified by this method.     

Computation of acoustic radiation from vibrating structures in motion

The noises from a vibrating structure in motion are often encountered in engineering practice, for example, tire noise and pass-by noise of moving vehicles. Consequently, research on the radiation characteristics of moving acoustic source is of significance. In this paper, a new computational method based on the wave superposition approach is developed for the acoustic field from a vibrating structure in motion. It inherits the advantages of the wave superposition approach in the acoustic computation, and in which a method of moving simple sources is used to eliminate the influence of the Doppler effect. By the proposed method, the acoustic radiation from the moving vibrating structure can be calculated easily with the same implementation process as the conventional wave superposition approach performed in the stationary acoustic field. Finally, the validity and feasibility of the proposed method are verified by the numerical results.     

The influence of turbulence on noise propagation from a point source above a flat ground surface

The presence of turbulence in the atmosphere affects the interaction between an acoustic wave and the ground surface. The noise attenuation by the ground in the presence of atmospheric turbulence is smaller than in non-turbulent atmosphere.A simple engineering model of noise propagation above a flat ground surface, for stationary and moving point sources, has been proposed. The model takes into account the air absorption and ground effect in the presence of turbulence.As well as parameters for type of ground and air absorption, the model introduces two adjustable parameters which must be deduced from in situ measurements at two ranges or two heights. The model’s free parameters have been obtained as a function of the resultant sound speed gradient on the basis of the field measurements performed for a stationary noise source. Also, using field data for a vehicle moving at steady speeds up to 100 km/h, the model has been verified for a moving point source.     

Method of control of the noise level of road vehicles in use

The elaboration of the method of control of noise generated by moving road vehicles has been made possible by using previous measurements of external noise produced by each of those particular vehicles. The results have made it possible to lay down the requirements for the design of specialised instrumentation capable of making such measurements. The advantage of this development is that it makes it possible for inexperienced technical staff to obtain the required acoustic measurements. The previously employed methods of noise prediction have required preceisely defined measuring conditions at the measuring site. The measurements, besides being laborious and time-consuming, give little information concerning the technical condition of the vehicle. On the contrary, with the method presented in this paper it is possible to measure not only the sound pressure level of the general external noise of the vehicle but also the band levels in the two frequency bands which are characteristic of the working engine at a definite speed of rotation, thus giving a preliminary estimation of the vehicle's noisiness. Defining the limiting values for noise level and band levels will enable excessively noisy vehicles to be eliminated.The results of measurements obtained with the above technique for various groups of vehicles are presented in this paper.     

Effects of speaker variability and noise on Mandarin fricative identification by native and non-native listeners

Speaker variability and noise are two common sources of acoustic variability. The goal of this study was to examine whether these two sources of acoustic variability affected native and non-native perception of Mandarin fricatives to different degrees. Multispeaker Mandarin fricative stimuli were presented to 40 native and 52 non-native listeners in two presentation formats (blocked by speaker and mixed across speakers). The stimuli were also mixed with speech-shaped noise to create five levels of signal-to- noise ratios. The results showed that noise affected non-native identification disproportionately. By contrast, the effect of speaker variability was comparable between the native and non-native listeners. Confusion patterns were interpreted with reference to the results of acoustic analysis, suggesting native and non-native listeners used distinct acoustic cues for fricative identification. It was concluded that not all sources of acoustic variability are treated equally by native and non-native listeners. Whereas noise compromised non-native fricative perception disproportionately, speaker variability did not pose a special challenge to the non-native listeners.     

基于智能视觉的车辆事故预警监控系统设计

针对车辆行驶中驾驶员视觉盲区或注意力不集中而造成的交通事故问题,提出采用智能视觉技术对车辆行驶过程中四周的异常物体进行视频监控,对动态视频场景中的运动目标进行检测、识别与实时测距,通过车辆智能视频监控系统最大程度的为驾驶员提供更多预警信息,预防交通事故的发生。本文介绍了车辆智能视觉监控系统的硬件设计方法与软件工作流程,并研究了运动目标检测算法与单目视觉测距算法,通过仿真实验,验证了本智能视觉车辆监控系统对于运动目标进行检测、识别与智能测距判断的实验结果。     

改进Fast-RCNN的双目视觉车辆检测方法

针对不同空间尺度的车辆表现出显著不同的特征导致检测算法效率低、准确性差且单目难以准确获取车辆距离信息的问题,提出了一种改进Fast-RCNN的汽车目标检测法,利用双目视觉对车辆进行测距。首先利用双目立体相机采集前方图像并进行预处理,加载深度神经网络Fast-RCNN的训练数据,再针对汽车不同空间尺度引入多个内置的子网络,将来自所有子网络的输出自适应组合对车辆进行检测,然后利用SURF特征匹配算法进行左右图像的立体匹配,根据匹配数据进行三维重建并确定车辆质心坐标,从而测量出车辆与双目相机之间的距离。实验结果表明,所述算法可以实现对车辆的快速检测,检测时间比传统的Fast-RCNN缩短了42 ms,并且实现了对5 m范围车辆距离的准确测量,其误差仅为2.4%,精确度高,实时性好。     

相干噪声快速去除算法及其在声源定位中的应用

针对相干噪声干扰声源辨识问题,将强跟踪滤波器理论与阵列-信号采集模型相结合,发展了一种快速估计相位变化的算法。算法引入多重次优渐消因子,能够进一步提取相位残差中的有用信息,使输出残差序列处处正交,且该因子能根据噪声相位差变化自动调节。通过仿真对连续相位突变进行跟踪表明,在参数失配条件下该算法实现了相位差的准确估计,且其性能在宽频范围具有稳定性。通过音箱实验给出了宽频范围内的成像结果,与已有算法对比表明,该算法不仅能够实时消除噪声干扰并可将收敛速度提高一倍以上。本研究实现了对随机变化相位的准确估计,提升了传声器阵列在相干噪声干扰下的声源实时定位能力。      

传声器阵列对高频弱声源的声成像识别方法

研究了传声器阵列对高频弱声源的识别定位方法。该方法根据高频声源的指向性和阵列探测特性等特点,提出了利用信噪比加权方法提高有效阵元对声成像的贡献,根据信噪比的大小对每个阵元添加不同的权值,可以显著提高传声器阵列对高频弱声源的声像清晰度。仿真分析了阵元加权和不加权两种方法对阵列声成像结果的影响,以某型号笔记本电脑电路板噪声为对象进行的实验表明,在阵列测量中充分利用有效阵元信号可以实现对声压级低达10~20dB的微弱噪声源的精确测量。      

燃料电池有轨电车声学优化*

基于线路噪声实验,系统测试分析了燃料电池有轨电车的噪声特性,研究了噪声分布以及空气传声、结构传声路径对噪声的贡献。结果表明改善车辆地板、空调、顶板和风挡的隔声性能,尤其是在500～1250 Hz的1/3倍频带范围内的隔声性能将有助于改善车辆内部声学环境。优化燃料电池系统控制,降低冷却单元转速将有助于改善车辆外部声学环境。在此基础上提出减震降噪建议措施,再次进行线路噪声实验,结果表明该措施有效。     

基于特征点光流聚类的复杂背景中运动车辆检测

为了准确、快速的在动态场景中对运动车辆进行检测,提出一种基于特征点光流聚类的车辆检测方法。该方法取Harris角点为特征量,通过对特征点做光流提取来剔除一些没有运动的干扰角点,然后再通过模糊U邻域(FUNN)聚类算法剔除噪音、孤立点和不感兴趣样本并实现前景和背景的分离,最后通过设定阈值判断前景目标是否是车辆。实验结果证明在复杂的动态场景中该算法具有更高的车辆识别率。     

Planar near-field acoustical holography in a moving medium

Near-field acoustical holography (NAH) is a well-established method to study acoustic radiation near a stationary sound source in a homogeneous, stationary medium. However, the current theory of NAH is not applicable to moving sound sources, such as automobiles and trains. In this paper, the inclusion of a moving medium (i.e., moving source and receiver) is introduced in the wave equation and a new set of equations for plannar NAH is developed. Equations are developed for the acoustic pressure, particle velocity, and intensity when mean flow is either parallel or perpendicular to the hologram plane. If the source and the measurement plane are moving at the same speed, the frequency Doppler effect is absent, but a wave number Doppler effect exists. This leads to errors when reconstructing the acoustic field both towards and away from the source using static NAH. To investigate these errors, a point source is studied analytically using planar NAH with flow in one direction. The effect of the medium moving parallel to the hologram plane is noted by a shift of the radiation circle in wave number space (k-space). A k-space Green's function and a k-space filter are developed that include the effects of the moving medium.     

Theoretical analysis and experimental research on non-cavitation noise modulation mechanism of underwater counter-rotation propeller

The underwater counter-rotation propeller non-cavitation noise has an obvious modulation characteristic which is due to the interaction of flow and blade. A modulation mechanism is presented in this paper. A sound pressure spectrum model is presented to describe its non-cavitation noise with application of generalized acoustic analogy method, the modulation mechanism is expressed with the improvement of sound pressure model. The power spectrum and modulation spectrum are presented by numerical simulation. Theoretical analysis and numerical simulation results are verified by the cavitation tunnel experiment. The modulation model of counter-rotation propeller is beneficial to the prediction modulation characteristics and identification of underwater high-speed vehicles.     

Influence of vehicle noise emission directivity on sound level distribution in a canyon street. Part I: Simulation program test

Simulation programs may be useful tools for controlling an environmental noise. The computer simulation program PROP11 that enables predictions of the time-average sound level within an urban system is used here. A roadway as a noise source is represented by a sum of the sound exposures due to individual vehicle pass-bys. Different representations of equivalent point sources for various classes of vehicles are allowed including directivity characteristics other than omnidirectional. Propagation throughout an urban system contains multi-reflections from the walls and single and double diffraction at their edges. In this paper, the PROP11 program is used to predict the sound level between opposing façades of buildings in a canyon street. A multi-lane road is assumed to contain two classes of vehicles (light and heavy) ones. The equivalent source representing vehicles is defined by the source power spectrum, its position above the ground and directivity characteristics. The consequences of introducing equivalent source directivity are analyzed.     

非紧凑低马赫数运动边界散射流动噪声的预测

提出了一种求解非紧凑低马赫数运动边界散射流动诱发噪声的预测方法。该方法首先基于运动坐标系下的连续性方程和动量方程推导得到该坐标系下的波动方程及其积分解,然后在该坐标系下采用边界元方法求解得到非紧凑运动边界表面的声压,最后将求解得到的壁面声压回代到静止或运动坐标系下的积分方程中实现对远场噪声的预测。推导得到的积分方程适用于流体与飞机机身、高速列车车身及旋转叶片等非紧凑结构边界作用诱发噪声的预测。     

基于声源预估的波叠加组合全息声源识别方法

为解决声源识别中旁瓣效应等因素引起的虚假识别结果问题,提出了基于声源位置预估的波叠加组合全息算法。应用波束形成算法进行声源位置的预估,将预估声压峰值点作为假定声源点,通过波叠加原理建立方程组,利用声源强度求解结果构造完整全息平面,通过声全息方法重新计算重建面声压。进行了仿真和实验验证,经该算法重构后,波束形成算法产生的虚假声源处的声压得以明显降低。表明该算法可以抑制旁瓣效应等因素的影响,提高声源识别的准确率。      

The identification of tyre induced vehicle interior noise

Sound transmission into a vehicle is classified as either airborne or structure borne sound. From the point view of noise control, the reduction of noise transferred by different paths requires different solutions. Coherence function analysis is often used to identify transmission paths. However it can be difficult to separate the airborne from structure borne components. The principle of acoustic reciprocity offers a convenient method for overcoming this difficulty. The principal states that the transfer function between an acoustic volume velocity source and an acoustic receiver is independent of a reversal of the position of source and receiver. The work done on this study involves exciting a stationary tyre and measuring the surface velocity of the tyre at a number of discrete points. The acoustic transfer functions between each point on the tyre and a receiver point are measured reciprocally. Two sets of measurements are then combined to yield a measure of the sound pressure due to a point force on the tyre via the acoustic transmission path only. This technique also provides information on the relative contributions of various regions of the tyre wall to the resultant noise. Also the sound radiation characteristics, the horn effect, and resonance at the wheel housing are identified through the reciprocal measurement.