采用不同小波母函数的阈值去噪方法性能分析

以用于测定汽油辛烷值的红外吸收光谱分析为背景,评估采用小波去噪方法时各种小波和阈值组合的去噪能力。文章构造了一个理想的原始光谱信号,考虑到小波去噪后信噪比以及原始光谱信号保留率这两者之间的协调关系,基于信噪比(SNR)定义了一个评价去噪优劣的评估系数η,在此基础上采用三种小波族系(Symlets,Daubechies,Coiflet)、四种阈值选取方法(Rigrsure,Sqtwolog,Heursure和Manimaxi)和三种阈值重调方法(One,Sln,Mln)对理想原始光谱信号进行了基于小波变换的信号去噪处理实验,以评价各种小波函数和阈值选取及重调方法的优劣。通过一系列的试验表明,对于该类型的信号,在实验所考察的小波族系和阈值选取及重调方法的范围内,采用Daubechies9或Symlet7,11,14,15小波,Rigrsure阈值选取规则和Sln阈值重调方法,可以得到最优的去噪性能。     

基于听觉模型的耳语音的声韵切分

本文分析了耳语音的特点，并根据生理声学及心理声学的基本理论与实验资料，提出了一种利用听觉模型来进行耳语音声韵切分的方法。这种适用于耳语音声韵切分的听觉感知模型主要分为四个层次：耳蜗对声音频率的分解机理；听觉系统的时域和频域非线性变化；中枢神经系统的侧抑制机理。这种模型能反映在噪声环境下人对低能量语音的听觉感知特性，因而适于耳语音识别，在耳语音声韵母切分实验中得到了满意的结果。     

基于听觉事件检测的汉语语音声韵切分

提出了一种基于听觉事件检测的汉语声韵母切分方法。该方法首先使用耳蜗滤波器组对语音进行滤波,然后在每个频带上检测对应于能量突变的听觉事件,最后在不同频率范围对听觉事件进行融合以确定声韵母边界。实验结果表明,对8 kHz采样的干净语音切分准确率可达到88.9%;信噪比10 dB的语音切分准确率可达到82.9%以上。     

基于熵函数的耳语音声韵分割法

耳语音声韵分割是耳语音识别和转换的前期工作。由于耳语发音不同于正常音,一般用于正常音的声韵分割法对耳语音不再适用。通过分析耳语音的发音及声学特点,利用宽带语谱图的声韵变化规律,提出了适用于耳语音的信息熵端点检测法,以及相对熵、音长和谱重心相结合的声韵分割法。并对两组信噪比为2-10 dB的380个汉语单音节耳语音进行声韵分割,女声的正确率为87.9%,男声的正确率为90.3%,高于频域法、聚类法和谱平坦度声韵分割法。实验表明,相对熵法可做为耳语音识别和转换的预处理,它改善了汉语耳语音转换为正常音的音质。     

调制转移函数和在室内声学中的应用

自从1973年在室内声学中引入调制转移函数(MTF)概念以来,实验研究表明MTF能够用来评价厅堂内语言传输的质量。发展了基于MTF概念的计算方法和专用测量仪器。MTF可使建筑师和声学工作者不论在设计阶段还是建成以后用来判断厅堂或扩声系统的语言可懂度。     

采用2D-Haar声学特征超向量的快速特定音频识别方法

针对特定音频事件识别技术在大数据音频处理任务中的准确性和快速性问题,提出一种基于2D-Haar声学特征超向量和AdaBoost算法的快速特定音频事件泛化识别方法。首先将多个连续音频帧的常用声学特征构成"声学特征图",进而提取维数高达数十万的Haar-like声学特征,然后使用AdaBoost.MH或速度较快的Random AdaBoost特征筛选算法,筛选出较高代表性的Haar-like声学特征模式组合,从而构成2D-Haar声学特征超向量;最后分析特定音频事件子类间的共性和差异性,提取子类别的共性,弱化子类间的差异,训练后得到一个泛化的音频事件模板,可支持多子类的泛化识别,能够准确检测并定位音频流中的特定音频事件。实验结果表明,使用2D-Haar声学特征超向量可以获得比MFCC,PLP,LPCC等常用声学特征约5%的识别精度提升、7~20倍的训练速度提升和5-10倍的识别速度提升,在网格法寻得最优参数配置下,可获得93.38%的准确率,95.03%的查全率,这为大数据量的特定音频事件识别提供了一种准确快速的处理方法。     

采用2D-Haar声学特征超向量的快速特定音频识别方法

针对特定音频事件识别技术在大数据音频处理任务中的准确性和快速性问题,提出一种基于2D-Haar声学特征超向量和AdaBoost算法的快速特定音频事件泛化识别方法。首先将多个连续音频帧的常用声学特征构成“声学特征图”,进而提取维数高达数十万的Haar-like声学特征,然后使用AdaBoost.MH或速度较快的Random AdaBoost特征筛选算法,筛选出较高代表性的Haar-like声学特征模式组合,从而构成2D-Haar声学特征超向量;最后分析特定音频事件子类间的共性和差异性,提取子类别的共性,弱化子类间的差异,训练后得到一个泛化的音频事件模板,可支持多子类的泛化识别,能够准确检测并定位音频流中的特定音频事件。实验结果表明,使用2D-Haar声学特征超向量可以获得比MFCC,PLP,LPCC等常用声学特征约5%的识别精度提升、7~20倍的训练速度提升和5-10倍的识别速度提升,在网格法寻得最优参数配置下,可获得93.38%的准确率,95.03%的查全率,这为大数据量的特定音频事件识别提供了一种准确快速的处理方法。     

基于频率特征的薄膜损伤声学判别方法

利用声学法研究单层氧化硅薄膜在不同激光能量下的损伤情况,建立了激光致薄膜产生的声波采集系统,对其采集到的时域信号进行傅里叶变换,比较和分析薄膜损伤前后声波24~40 kHz高频段的曲线,提取频率特征,并提出用曲线相似函数进行薄膜损伤的识别方法。实验数据结果表明,该方法简单易行,可降低环境噪声的影响,既能实现在线检测,又能准确判别薄膜损伤。     

基于频率特征的薄膜损伤声学判别方法

利用声学法研究单层氧化硅薄膜在不同激光能量下的损伤情况,建立了激光致薄膜产生的声波采集系统,对其采集到的时域信号进行傅里叶变换,比较和分析薄膜损伤前后声波24~40kHz高频段的曲线,提取频率特征,并提出用曲线相似函数进行薄膜损伤的识别方法。实验数据结果表明,该方法简单易行,可降低环境噪声的影响,既能实现在线检测,又能准确判别薄膜损伤。     

采用拖曳线列阵的海洋声学参数联合反演方法研究

匹配场反演方法是快速获取海底声学参数的一种有效途径,但是其反映的是水体和海底空间变化环境的平均效果,对海底密度和衰减系数的敏感性较小,且在复杂海底环境下,不确定性明显增加。匹配场反演方法通常采用垂直阵来实现,其机动性较差,只能获得局部海区的环境参数。利用拖曳线列阵可以实现环境参数的走航式获取。提出了基于匹配场与反射损失联合反演的方法,利用其中之一的反演结果作为下一步反演的先验信息,敏感参数与不敏感参数多步反演,并逐步缩小了参数的优化空间。同时引入贝叶斯理论与蒙特卡罗方法对反演参数的后验概率密度进行分析,并利用反演结果的后验分布计算了传播损失随距离的概率分布。研究结果表明,联合反演方法对地声参数进行了更准确的反演,尤其是对海底声速反演效果提高明显。通过传播损失的概率分布发现,估计得到的海底参数用于声场预报时,误差较小。     

基于鼻韵尾分离的汉语声韵母识别模型

研究汉语自然口语识别中的建模单元选择问题。在HMM三状态模型中,声韵母单元与音素单元作为两种最流行的建模单元各有优劣。一方面从自然口语音变严重的问题出发,倾向采用粗粒度的声韵母单元以概括各种音变;另一方面从三状态结构可能无法有效描述复杂单元的问题出发,又倾向采用细粒度的音素单元。本文在实验语音学理论研究成果与声韵母时长分析实验结果的基础上,主张对扩展声韵母单元进行有选择的拆分,提出了基于鼻韵尾分离的声韵母拆分方法。实验结果表明本文的方法与扩展声韵母单元、音素单元相比,识别性能有了明显改善,其字错误率分别降低2.23%和9.45%。     

Spatio-temporal segmentation of moving objects using edge features in infrared videos

In this paper, a novel method is proposed for spatio-temporal segmentation of moving objects using edge features in infrared videos. We define motion saliency of edge (MSoE) to generate the MSoE-map. The seeds of moving objects are extracted from the MSoE-map by using Otsu's method and subsequently compensated by historical data. An improved layer-based region growing method is applied to the seeds to achieve spatial segmentation of moving objects. The region growing method has an adjustable growing threshold. So, one of the focuses of our work is how to determine the best growing threshold. A Markov Random Field (MRF) based criterion with maximum a posterior (MAP) estimation principle is proposed for performance evaluation of moving object segmentation without ground truth (GT) in infrared videos. This criterion can be considered as an object function of threshold determination during global searching. The global optimum is accomplished by using simulated annealing (SA) algorithm to obtain the best growing threshold. The final segmentation mask of moving objects is grown from the seeds with the best growing threshold. Experimental results are provided to illustrate that the proposed method has better performance for moving object segmentation with fewer effects of object-background misclassification in infrared videos.     

Insertion loss testing of active noise reduction headsets using acoustic fixture

Active Noise Reduction (ANR) technique has been available for hearing protectors and communication headsets for years. However, no standard testing method for ANR headsets has emerged. In this study, we measured insertion losses of four types of commercially available ANR headsets using an Acoustic Test Fixture (ATF) to examine the feasibility of using ATF for this kind of measurement. The measurement outlined required a minimum of instrumentation and was relatively simple to implement. The results show that this method should be especially useful for testing of prototypes and for quality control as well.     

Transmission loss measurement of acoustic material using time-domain pulse-separation method (L)

An alternative method for measuring the normal incidence sound transmission loss (nSTL) is presented in this paper based on the time-domain separation of so-called Butterworth pulse with a short-duration time about 1 ms in a standing wave tube. During the generation process of the pulse, inverse filter principle was adopted to compensate the loudspeaker response, besides this, the effect of the characteristics of tube termination can be eliminated through the generation process of the pulse so as to obtain a single plane pulse wave in the standing wave tube which makes the nSTL measurement very simple. A polyurethane foam material with low transmission loss and a kind of rubber material with relatively high transmission loss are used to verify the proposed method. When compared with the traditional two-load method, a relatively good agreement between these two methods can be observed. The main error of this method results from the measuring accuracy of the amplitude of transmission coefficient.     

骨自动分割及基于轮廓一致性的分割验证

利用序列CT图像对骨组织做自动分割是计算机辅助骨科的重要技术。由于骨结构的不一致性、病理改变以及CT数据内在的模糊性,完全做到自动分割目前还面临着很大的困难。提供了一种有效的解决框架,既避免了传统分割中大量的人力介入,又避免了自动分割造成的错误。具体步骤是:设计合适的自动分割算法对序列CT图像中骨组织做自动分割;利用相邻CT图像上骨轮廓是渐进变化的特性,通过比较轮廓形状的一致性来自动验证分割的结果;对一致性较差的结果引入决策判别机制。     

Optimisation using measured Green's function for improving spatial coherence in acoustic measurements

Aberrating materials can degrade acoustic measurements by distorting the acoustic wavefront and causing acoustic speckle (as opposed to speckle noise which is a manifestation of coherent backscatter). The amplitude and phase fluctuations associated with acoustic speckle can introduce considerable measurement uncertainty which is difficult to deal with. This paper demonstrates a new technique which optimises the spatial distribution of the generation of the ultrasound to compensate for the aberration. This technique uses experimentally measured Green's functions to allow the calculation of the field resulting from the generation wavefront during optimisation. The technique is used to improve the accuracy of velocity measurements in a steel sample using 82 MHz SAW waves. This is achieved by optimising for improved spatial coherence in the measurement region which suppresses the speckle noise. Experimental evidence of acoustic aberration arising from grain structure is shown for steel and aluminium and the measured Green's function optimisation technique is shown to overcome the resulting acoustic speckle. The technique was performed using the Adaptive Optical Scanning Acoustic Microscope (AOSAM) at Nottingham University, UK.     

Determination of a response function of a thermocouple using a short acoustic pulse

This paper reports on an experimental technique to determine a response function of a thermocouple using a short acoustic pulse wave. A pulse of 10 ms is generated in a tube filled with 1 bar helium gas. The temperature is measured using the thermocouple. The reference temperature is deduced from the measured pressure on the basis of a laminar oscillating flow theory. The response function of the thermocouple is obtained as a function of frequency below 50 Hz through a comparison between the measured and reference temperatures.