多通道噪声测量的关键理论与应用研究

发布时间：2018-03-28 01:10

本文选题：多通道噪声测量　切入点：自卷积检测　出处：《湖南大学》2015年博士论文

【摘要】：噪声是一种人们不希望听到的声音,对人类的生活和健康产生巨大的影响,对动物、仪器设备和建筑物等各方面也产生一定的影响。噪声污染同水污染、大气污染和废物污染一起被看成是世界范围内的四个主要环境问题。随着社会经济的发展,噪声污染问题越来越突出,带来了一系列的不良效应,影响社会的和谐稳定发展。因此,开展噪声测量与分析技术的研究具有非常重要的意义。客观而全面的噪声测量与分析是认识、判断和处理各种噪声问题的重要手段。随着科技的发展以及计算机技术水平的提高,传统噪声测量仪器的功能和性能等发生了很大的变化,正向数字化、智能化、网络化和虚拟化的方向发展。但是国内外高性能噪声测量仪器和系统价格比较昂贵,而一般的噪声测量仪器和系统存在各种各样的不足。本文采用理论推导、仿真分析和实验验证相结合的方法,拟设计和开发多通道噪声测量与分析系统,并且针对多通道噪声测量与分析相关的关键理论开展深入的研究。主要有以下几个方面:(1)在多通道噪声测量过程中,由于各种因素的影响,尤其是背景噪声的影响,每个通道测量得到的信号是不能直接被使用和处理的,否则会影响测量和评价的结果,因此针对噪声测量的预处理问题,研究周期信号的自卷积去噪理论,提出自卷积检测法,拟用于去除噪声测量中的背景噪声,提高被处理信号的信噪比。对各种不同频率成分的信号进行仿真实验。(2)在多通道噪声测量过程中,由于声音无处不在,所以传声器接收到的信号一般是多个噪声源信号的混合,而如果要测量单个噪声源信号,则可以先对混合多噪声源进行分离。针对混响环境中多个噪声源同时存在时的声源分离和单个噪声源实际辐射噪声的测量问题,研究基于时间反转技术(Time reversal technique,TRT)的声源分离和测量方法,并在MATLAB环境中对混响室内的二维和三维声场进行建模仿真,讨论不同的声场参数,如声源位置、声源类型、麦克风阵列布局、混响和通道噪声对算法性能的影响。(3)在多通道噪声测量与分析中,有些应用场合需要对每个通道测量的信号进行频谱分析,分析噪声的频率成分或频谱特征。由于快速傅里叶变换(Fast Fourier transform,FFT)算法本身的缺陷,对频谱特征的提取需要进行离散频谱校正,如估计质心频率。针对传统的离散频谱校正方法的不足,研究基于谱质心(Spectral centroid,SC)的声信号频谱校正方法。讨论SC应用于单频信号和多频信号的频谱校正原理,尤其针对常见的声信号估计宽频和倍频程的SC及质心频率,并且讨论在背景噪声下SC的变化。(4)由于在多通道噪声测量与分析系统中,每个通道测量一路噪声信号,且每个通道的数据是独立处理和分析的,所以研究单通道噪声测量数据的后处理方法仍然非常重要。针对单通道噪声测量数据的信息融合和拟合问题,提出基于信息量和最小条件熵参数估计两种信息融合方法以及基于互信息的数据拟合辨识方法。对于信息量融合方法,首先利用最大熵方法(Maximum entropy method,MEM)估计测量样本的概率分布,再根据每个样本的自信息量与样本总体信息熵的比值对样本数据进行融合;对于最小条件熵参数估计融合方法,根据观测样本的条件概率密度函数构建观测总体条件信息熵,最小化该条件信息熵,求解无约束极值问题即可得到最优结果;对于互信息拟合辨识方法,将数据拟合的过程看作是一个通信过程,根据信息通信的信道模型构建数据拟合模型,再根据拟合数据、拟合曲线和拟合误差三者的信息熵求解拟合模型的互信息,选取互信息最大的曲线作为拟合曲线。(5)结合多通道噪声测量与分析系统的需求和性能指标,以计算机为信息处理核心,结合虚拟仪器、数据库、高速数据采集卡和数字信号处理等多种应用技术设计开发一款多通道噪声测量与分析系统,拟达到通道多、功能强、精度高、速度快、价格低等性能指标,详细给出每个软件模块的设计原理和程序。
[Abstract]:Noise is a kind of unwanted sound, have a tremendous impact on human life and health of the animal, the equipment and buildings also have a certain impact. Noise pollution and water pollution, air pollution and waste pollution is regarded as the world within the scope of the four major environmental problems. With the development of social economy, the problem of noise pollution is becoming more and more prominent, brought a series of negative effects, affect social harmony and stability and development. Therefore, it is very important to study the development of noise measurement and analysis technology. Noise measurement and analysis of objective and comprehensive understanding, an important means to judge and deal with various noise problems. With the development of technology and improve the level of computer technology, has undergone great changes, the traditional noise measuring instrument function and performance are digital, intelligent, network The virtual and the direction of development. But the domestic and foreign high performance noise measuring instruments and systems are expensive, lack of noise measuring instruments and systems in general there are various. This paper uses the method of theoretical derivation, simulation analysis and experimental validation of the combination, to the design and development of multi channel noise measurement and analysis system, and for deep research of the key theory and related analysis of multi channel noise measurement. Mainly in the following aspects: (1) in multi channel noise in the measurement process, due to various factors, especially the influence of background noise, signal of each channel measurement is not to be used directly, otherwise it will affect the measurement and the results of the evaluation, so the pretreatment problem of noise measurement, self convolution denoising theory of periodic signal, the self convolution method, to be used for removing noise measurement In the background noise, improve signal processing by SNR. Signals of different frequencies is simulated. (2) in multi channel noise in the measurement process, the sound is everywhere, so signal received by the microphones are generally mixed multiple noise source signals, and if you want to measure a single noise source signal you can first, separating the mixed noise sources. To solve the problem of measuring the sound source separation and single noise source of multiple noise sources in reverberant environments exist at the same time the actual radiation noise, the research based on time reversal technique (Time reversal technique, TRT) of the sound source separation and measurement method, and two-dimensional in MATLAB environment the reverberation chamber and the three-dimensional acoustic modeling and simulation, discuss the acoustic parameters, such as the position of the sound source, the sound source type, microphone array layout, and the effects of reverberation channel noise on the performance of the algorithm in (3). Measurement and analysis of noise in the channel, some applications need to analysis the signal spectrum of each channel measurement and analysis of noise frequency components or spectrum characteristics. Because of the fast Fourier transform (Fast Fourier transform FFT) algorithm to extract the defect itself, the need for spectrum correction of discrete spectrum estimation, such as centroid frequency for. Lack of correction method for discrete spectrum of the traditional research based on spectral centroid (Spectral centroid, SC) of the sound signal spectrum correction method. Discuss the correction principle of spectrum of SC application in single frequency and multi frequency signal, especially for SC and centroid frequency broadband and octave estimation of acoustic signals in common, and discuss the changes in the background the noise of SC. (4) due to the multi channel noise measurement and analysis system, each channel measuring a noise signal, and each channel is independent of the data processing and analysis So, study on single channel noise measurement data postprocessing method is still very important. Based on the information fusion of single channel noise measurement and data fitting problem, based on the two kinds of information fusion method of information quantity and the minimum conditional entropy of parameter estimation and identification data fitting method based on mutual information. The information fusion method based on the maximum. Entropy method (Maximum entropy method, MEM) to estimate the probability distribution of the measured samples, then according to the ratio of each sample from the amount of information and the sample information entropy fusion of sample data; the small conditional entropy fusion method for parameter estimation, according to the observation samples of the conditional probability density function to construct general observation conditional information entropy, minimizing the conditional information entropy, solve the unconstrained extremum problem can get best results; mutual information for fitting the identification method, the process of data fitting at As a communication process, according to the construction of data fitting model of channel model of information communication, according to the fitting data, mutual information entropy fitting model fitting curve and the fitting error of the three, select the maximum mutual information as curve fitting curve. (5) according to the requirements and performance index of multi channel noise measurement and analysis the system, based on computer information processing core, combined with virtual instrument, database, the development of a multi channel noise measurement and analysis system of high speed data acquisition card and a variety of applications such as digital signal processing technology to design, to multi-channel, strong function, high precision, fast speed, low price performance index, design principle and procedure details are given of each software module.

【学位授予单位】：湖南大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TB53

【相似文献】