QQ客服 官方微博 反馈留言

题名	随机共振理论及其在信号处理中的应用研究
作者	叶青华
学位类别	博士
答辩日期	2005
授予单位	中国科学院声学研究所
授予地点	中国科学院声学研究所
关键词	随机共振 微弱信号检测 阵列信号处理 微弱图像增强 模数转化
中文摘要	随机共振是非线性系统、随机噪声和输入信号之间产生的一种奇特的协同现象，它反映了噪声的积极作用。本文在水声信号处理的背景下，从随机共振物理机理与信号处理的角度，从理论、信号检测、阵列处理、图像增强等四个方面研究了随机共振理论及其在信号处理中的应用，这些研究对于随机共振理论和应用的进一步发展，以及对我们的主要工作一水声信号处理都很有意义。为了使随机共振理论走向实用化，本文在随机共振系统的参数最优化方面作了大量深入的研究工作。分析了双稳态系统最优输出时系统参数间关系，据此得到了一种调节系统参数的简单方法；针对一般的有记忆系统只能处理单频或窄带信号，，并且对低频信号增强更有效的情况，我们将高频或宽带信号子带分解后移频至基带，能够很好地解决这个问题；建立了微弱图像增强的性能度量方法，并以此为基础，提出了基于噪声样本集合的参数选择算法，并能够推广到其它的随机共振系统和信号类型。本文对随机共振检测器进行了深入的探讨，对高斯噪声、非高斯噪声下的最优线性、最优非线性以及随机共振检测器作了详细的分析。得到以下结论：（1）在检测性能上，高斯噪声中随机共振检测器接近于匹配滤波检测器；非高斯噪声中，随机共振检测器优于匹配滤波检测器；（2）在实现上，非高斯噪声中的最优非线性检测器结构复杂，很难实现；而性能次优的随机共振检测器实现非常简单。研究了采用阵列随机共振系统的信号检测问题，发现在很多噪声类型下，随机共振阵列检测器的检测性能接近于最优检测器（即Neyman-Pearson准则下的非高斯噪声中的最优非线性检测器，和高斯噪声中的最优线性检测器）。因此，在非高斯噪声中的信号检测问题中，我们不需要构造复杂的非线性检测器，采取简单的随机共振阵列检测器就能接近最优的检测性能。另外，不需要知道精确的噪声概率密度函数，这是一种宽容的信号处理方法。在阵列信号处理中，本文研究了将随机共振作为前置处理器的阵列处理、高频和宽带阵列信号的处理、基于随机共振的时一空处理方法、小信号在数字波束形成中的随机共振效应、背景噪声对阵列信号检测能力的增强作用、方位估计中另加噪声的积极作用，以及随机共振在实时阵列信号模拟软件中的应用。提出了随机共振A／D转换的概念和方法，并和常规的A／D转换进行了比较。研究表明，不仅对于阵列信号检测，在目标方位·估计中噪声也能提高处理性能。基于随机共振系统的阵列处理不仅简化了系统，而且性能上基本等效于常规方法，在非高斯噪声下优于常规方法。另外，基于比特量化的信号模拟也可以应用于实时阵列信号模拟软件。在图像增强领域，本文建立了基于随机共振的一套比较完善的理论和方法。将有记忆系统应用于图像中直流信号的增强，扩展了有记忆随机共振系统的应用范围，并且采用线、邻域、块等增强算法，与调整系统参数方法结合，可以处理一些极低信噪比的图像增强问题。
英文摘要	Stochastic resonance (SR) is the phenomenon of enhancement of signal transmission by certain nonlinear systems resulting from the addition of noise to the system. In the view of SR mechanism and signal processing, theory and applications of SR are studied in this thesis from four aspects, i.e. theory, signal detection, array processing and image enhancement. It is important for the development of SR theory and applications and for underwater acoustic signal processing. For practical application, much work has been done on the system optimization. A) It is assumed that the signal and noise properties are known in most bistable SR studies. To overcome this shortcoming, the relationship among the parameters of a bistable SR system is investigated and a system can be moved into the SR condition by conveniently changing its structural parameters. B) Conventional memory SR systems can only enhance narrowband and low frequency signals. By decomposing the input signal into subbands and translating the subband signals to baseband, a high frequency signal can also be enhanced. C) A weak image enhancement performance measure is put forward and a parameter optimization algorithm based on noise sample set is presented. This algorithm can be extended to other SR systems and other types of signals. Signal detection based on SR is discussed, and optimum linear, optimum nonlinear, and SR detector in Gaussian or non-Gaussian noise are analyzed in detail. Conclusions can be obtained: A) For signal detection performance, SR detector nearly comes up to matched filter detector in Gaussian noise, and performs better than matched filter detector in non-Gaussian noise. B) For system structure, the optimum nonlinear detector in non-Gaussian noise is complicated, while the SR detector is simple to implement. Detection based on SR array is also studied and it is found that the performance of SR array detector approaches that of optimum detector (i.e. optimum nonlinear detector in non-Gaussian noise and optimum linear detector in Gaussian noise) in many different noise densities. Then it's no need to construct complicated nonlinear detector in non-Gaussian noise and the SR array detector can nearly achieve the best performance. In addition, this is a robust detection method because we don't need to know the exact noise probability density function. Array processing using SR as preprocessor, wideband and high frequency array signal processing using SR, SR based time-space processing, SR effect of weak signal in digital beamforming, the constructive role of background noise in array signal detection, noise enhanced performance in direction of arrival (DOA) estimation and the application to real-time array signal software simulator are studied in this thesis. The concept and method of SR A/D converter are presented and compared to the conventional A/D converter. It is found that noise can improve the system performance in array signal detection and in DOA estimation. Using SR as the basic processing method not only makes the system structure simple to implement, but also keeps nearly the same performance to conventional methods in Gaussian noise while improves the performance in non-Gaussian noise. The signal simulation method based on one-bit quantization can also applied in real-time array signal simulation software. In addition, a framework of image enhancement using SR is formed in this thesis. The applications of memory SR systems are expanded to image enhancement, and the line, circle and region enhancement algorithms are presented to enhance low signal-to-noise ratio images.
语种	中文
公开日期	2011-05-07
页码	128
内容类型	学位论文
源URL	[http://159.226.59.140/handle/311008/892]
专题	声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文
推荐引用方式 GB/T 7714	叶青华. 随机共振理论及其在信号处理中的应用研究[D]. 中国科学院声学研究所. 中国科学院声学研究所. 2005.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。