Research On Critical Theories And Applications Of Multi-Channel Noise Measurement

Noise is a kind of sound that people do not want to hear and has huge impact on human life and health, also some impact on animal, instrument and equipment, building, et al. Noise pollution, water pollution, air pollution, and waste pollution are seen as four major environmental problems within worldwide. With the development of society and economy, noise pollution is becoming increasingly prominent and has brought a series of adverse effects, which influences the development of society harmoniously and stably. So the research on noise measuring and analyzing techniques has very important significance. Objective and general noise measurement and analysis are important means to acquaint, judge, and process various noise problems.With the development of science and technology and improvement of computer technology, traditional noise measuring instruments’ functions and performances have great changes and are developing to digital, intelligent, networking, and virtualization directions. But the prices of noise measuring instruments and systems with high performance are expensive at domestic and foreign markets, while general noise measuring instruments and systems have many shortages.Combing theory derivations with simulation analyses and experimental verifications, the thesis is to design and develop multi-channel noise measuring and analyzing system(MNMAS) and carry out depth research on some critical technologies related to multi-channel noise measurement and analysis. Main work is as followed:Firstly, in the process of multi-channel noise measurement, each channel’s signal cannot be directly used and processed for various factors’ influence, especially background noise; otherwise it will affect the results of measurement and evaluation. About the pre-processing problems of noise measurement, the de-noising theory of periodic signal’s self-convolution is studied and a method of self-convolution detection is proposed to eliminate background noise in noise measurement and improve the processed signals’ SNR. The thesis carries out some simulation experiments using some signals with different frequency contents.Secondly, in the process of multi-channel noise measurement, for sound disperses everywhere, the received signal in a microphone is a mixed-signal by multiple sound sources. If you want to measure a single sound source signal, then you must firstly separate the mixed-signal. About the separation of sound sources and measurement of sound actually radiated from a single sound source when multiple sound sources are present under reverberation environment, a method of sound sources separation and measurement based on TRT is proposed. In MATLAB, numerical simulations are carried out through building 2D and 3D models of reverberation room. The effects of different sound field parameters, including of sound source’s position, sound source’s type, microphone array’s layout, reverberation, and channel noise on the performance of the proposed algorithm are discussed.Thirdly, in the process of multi-channel noise measurement and analysis, some applications require spectrum analysis of measured signal in one channel and to analyze noise’s frequency components or spectral features. For FFT’s defects itself, extracting spectral features needs discrete spectrum c orrection, such as estimating centroid frequency. About the shortages of traditional spectrum correction methods, a method of spectrum correction for acoustic signal is proposed based on SC. The correction theories of SC for single-frequency signal and multi-frequency signal are discussed. Especiall y for normal acoustic signals, the thesis estimates their SCs and centroid frequencies of wide frequency and octave, also discusses their changes under background noise.Fourthly, because every channel measures one noise signal and every channel’s data is processed and analyzed independently in MNMAS, the research on post-processing methods for single-channel noise measuring data is still very important. About the information fusion and fitting of single-channel noise measuring data, two methods of information fusion based on information content and minimum conditional entropy parameter estimation and a recognition method of data fitting based on mutual information are proposed. For the information content fusion method, after estimating the samples’ probability distribution using MEM, the samples are fused according to the ratio of every sample’s information content and samples’ total entropy; for the fusion method of minimum conditional entropy parameter estimation, the observed total conditional entropy is build according to the conditional probability density function of observed samples and is minimized to get the op timal result through solving unconstrained optimization problem; for the data fitting recognition method of mutual information, the fitting course is looked as a communication course and a data fitting model is build based on information communication’s channel model. The mutual information of fitting model is calculated by three entropies of fitting data, fitting curve, and fitting error. That curve with biggest mutual information is chosen as the best fitting curve.Lastly, combining MNMAS’s demands and performance indicators, MNMAS is designed and developed taking a computer as a kernel of information processing and combining virtual instrument, database, high-speed data acquisition card, and di gital signal processing, et al and the system has multiple channels, strong function, high accuracy, fast speed, and low price. The thesis provides the design principle and program of every software module.