【作者】 于凤芹；

【导师】 曹家麟；

【作者基本信息】 上海大学 ， 控制理论与控制工程， 2005， 博士

【摘要】 Chirp信号是一种典型的非平稳信号,广泛地应用于通信、雷达、声纳、生物医学和地震勘探等系统,许多自然现象也可以用Chirp信号作为信号模型。本文以多分量Chirp信号为研究对象,采用时频重排-Hough变换、分数阶傅立叶变换、三参数Chirp原子分解等方法,对多分量Chirp信号的时频表示、检测与参数估计进行了较为深入的研究。 从时频聚集性、交叉项抑制、分量间的可分辨性等方面,研究了模糊函数、WVD及Cohen类双线性时频分布对多分量Chirp信号的时频表示性能。尽管模糊函数和WVD对单分量Chirp信号有最好的时频聚集性,但它们对多分量Chirp信号的时频表示却存在交叉项干扰。Cohen类时频分布使用核函数平滑来减少交叉项,但平滑也使时频聚集性下降,交叉项和较低的时频聚集性又导致分量间的可分辨性变差,所以,对于多分量Chirp信号,现有的双线性时频分布不是最好的时频分析工具。 由于多分量Chirp信号在时频平面淹没在交叉项和噪声中,直接从其时频分布很难辨别信号的存在。为此提出了将该信号的双线性时频分布作为图像,由于Chirp信号在时频平面呈直线,利用Hough变换检测图像中直线的原理,将噪声中多分量Chirp信号的检测与参数估计转换为在参数空间寻找局部极大值及其相应坐标的问题,使检测和参数估计一并完成。针对WVD中的交叉项经Hough变换后会形成伪尖峰;抑制此伪尖峰而平滑交叉项又造成时频聚集性下降这一矛盾,提出了基于时频重排-Hough变换和抽取图像脊-Hough变换的两种改进方法,从而既解决了交叉项抑制与时频聚集性下降的矛盾,又缩短了Hough变换的时间,提高了检测和参数估计的准确性,也改善了抗噪声性能,并将时频重排-Hough变换方法应用于对噪声中多运动目标雷达回波信号的检测,仿真实验结果证明了该方法的有效性。 深入系统地研究了分数阶傅立叶变换(FRFT:Fractional Fourier Transform)的基本理论,包括傅立叶变换算子的分数阶化过程、FRFT的定义与性质、典型信号的FRFT、分数阶傅立叶域的概念和运算、FRFT与其它的时频表示的关系、以及离散FRFT的计算等。我们的目的是从多个角度揭示和分析多分量Chirp信号在适当的分数阶傅立叶变换域呈现冲激信号特征的机理。利用这个特征,提出了对直接序列扩频通信系统中多个Chirp宽带干扰的识别和剔除方法,仿真实验结果说明了这利方法的可行性。 研究了Chirplet变换的定义和物理意义,它适合于对多分量Chirp信号的线更多还原

【Abstract】 Chirp signal is a typical non-stationary signal, which widely appears in systems such as communication, radar, sonar, biomedicine and earthquake exploring system and also be used as signal model for a lot of natural phenomena. The detection, the parameters estimation and the time-frequency representation for the multi-component Chirps were researched in this dissertation, using the reassignment time-frequency distribution-Hough transform (RTFD-HT), fractional Fourier transform (FRFT), and tree-parameter Chirp atom decomposition methods.The time-frequency distribution performance of ambiguity function, WVD, and Cohen class time-frequency distribution for the multi-component Chirps was described from the time-frequency concentration, the cross terms reduction, and the discriminability. Though WVD and ambiguity function is a better time-frequency representation for mono-component Chirp, the interferences of cross terms of the bilinear time-frequency distributions occurred for the multi-component Chirps. Using kernel function smooth can restrain the cross terms, but it results in a deterioration of the time-frequency concentration and the discriminability. Therefore, the Cohen class bilinear time-frequency distributions are not the best as a time-frequency analysis tool for the multi-component Chirps.Because the presence of the multi-component Chirps is barely identified directly from time-frequency distribution due to the existence of the cross terms and noise, the method of the detection and the parameters estimation for the multi-component Chirps was proposed, based on time-frequency distribution-Hough transform, which regards the bilinear time-frequency distribution of the signal as an image, applies the principle of the detecting line of Hough transform and finds the local maximal and relevant coordinate in order to detect the line and estimate its parameter. Confronting with the contradiction between restraining the cross terms and increasing the time-frequency concentration, we brought forward the improved methods of the reassignment time-frequency -HT and the image ridges extraction-HT. The simulation results have shown that the proposed method can not only eliminate false detection resulted from pretense peak value of the cross-term and shorten the time of Hough transform, but also increase the accuracy of the detection and parameters estimation and reduce noisy interference. Also the proposed method has been applied to detect a multi-target radar signal in noise and has, through simulation更多还原