【作者】 郭建涛；

【导师】 王宏远；

【作者基本信息】 华中科技大学 ， 信号与信息处理， 2009， 博士

【摘要】 跳频是最常用的扩频方式之一，跳频通信具有良好的抗干扰性，低截获概率和易于组网的特点，因此跳频技术一出现，便在军事领域得到了极大的发展。随着跳频技术及各种硬件技术的发展，跳频通信向着高跳速、宽跳频带和多跳频点数的趋势发展，并且逐渐与其它扩频技术相结合，采用自适应跳频的工作方式，极大地提高了军事装备的抗截获和抗干扰能力，特别是在现代电子战密集的信号环境下，向通信对抗提出了严重的挑战。时频分析作为一种新兴的信号处理方法，可以克服传统傅立叶变换在处理非平稳信号时的局限，用时间和频率的二维函数描述出信号频率随时间的变化规律。近年来，它在理论研究方面取得了重大进展，并在实践中得到了广泛的应用。通过时频分析方法得到混杂着噪声的非合作跳频信号的时频分布，估计其跳频速率、跳变时刻和跳频频率等参数，是截获敌方通信、产生最佳干扰信号、瓦解敌方正常通信的前提，因此成为现代军事通信对抗中的研究重点之一。本文主要讨论了提高跳频信号时频分布的聚集性和交叉项抑制性能的核函数设计方法和熵测度评价准则，以及基于粒子群算法的跳频信号自适应分解等问题，对跳频信号的参数估计算法进行了广泛深入的研究。本文首先推导了跳频信号的模糊函数公式，利用跳频信号魏格纳分布和模糊函数的关系，以及其模糊函数自项集中在模糊平面原点，而交叉项远离原点位置的属性，设计了一种具有抽样函数形状的核函数。由于设计核函数实现了与跳频信号模糊函数自项更高程度的匹配，因此相应时频分布能够有效抑制交叉项的干扰，保留尽可能多的自项能量，获得更好的参数估计性能。Cohen类时频分布不仅类型众多，而且都包含一个或多个参数，因此核函数的形状和参数的选择对时频分布的性能影响较大。如何评价时频分布的优劣，选取合适的时频分布类型及其参数，成为跳频信号时频分析一个主要问题。本文从交叉项抑制和聚集性提高两个角度，分析了熵测度在时频分布性能评价中的反映，提出了一种最小熵测度准则和SQP算法相结合的核参数优化方法，相应的时频分析方法有效降低了跳频信号参数估计的方差，提高了参数估计的精度。从实现的方法来划分，时频分析主要分为原子分解和能量分布两大类。原子分解把信号分解为在时间和频率上都有明确意义的时频原子的线性组合，获取的时频分布从根本上消除了交叉项的干扰。为了克服匹配追踪算法在信号原子分解时运算量大的瓶颈，本文引入了粒子群优化算法。在获取与跳频信号分量匹配的原子参数的基础上，设计了一种跳频信号参数估计的新算法。该算法摆脱了基于时频平面参数估计的信噪比阈值效应，能够在低信噪比的条件下获得良好的估计效果。在基于匹配追踪的粒子群算法中，每次迭代只能选取跳频信号的单个分量，降低了算法的运行效率。利用信号与过完备库原子内积的多峰性，提出了基于物种形成粒子群优化的参数估计算法，充分利用了粒子群在搜索过程中的收敛性。计算机仿真结果表明，该算法在一次运行过程中能够给出所有与跳频信号分量相匹配的时频原子，适合多种跳频信号的参数估计。更多还原

【Abstract】 Frequency Hopping (FH) is one of the most commonly used way about spread spectrum communication technique which develops fast in military field due to its good immunity against interferences, low probability of intercept and facility in communication networking. Along with the development of FH and each kind of hardware technology, the FH communications turn toward high hop rate, wide hop frequency band and more FH points. Farther, FH technology is unified with other spread spectrum ones gradually and uses the auto-adapted FH working which enhanced the ability of military equipments against jamming and interception enormously and brought up a new austere challenge for communication countermeasures, especially under crowed signal environment in the modern electronic warfare.As a novel signal processing method, time-frequency (TF) analysis may overcome the drawback of Fourier Transform about nonstationary signal process, which could depict the frequency change law of signal versus time using the two-dimensional function of time and frequency. In recent years, it has made the significant development in the fundamental research aspect, and obtained the widespread application in reality. Obtaining TF representation of the non-cooperation FH signal embedded in noise and its parameter such as hopping during, hop timing and hopping frequency is precondition to achieve the interception enemy side correspondence, produce the best jamming signal and disintegrate the enemy side normal correspondence finally, therefore becomes in modern military correspondence countermeasure research key one. In the present paper, the problems about kernel function design and entropy measure appraisal to enhance the TF concentration and cross-term suppression of TF distribution of FH signal, and atomic decomposition based on particle swarm optimization are discussed. At last, this thesis has engaged in extensive research on parameter estimation algorithm about FH signal.In this thesis, the FH signal ambiguity function formula has been deduced firstly, which the auto-components are centered the origin of the ambiguity plane and the cross-components are located away from it, and derived one kernel function in sampling function shape using the relations of Wigner distribution and ambiguity function of FH signal. The corresponding TF distribution could suppress the cross-term interference effectively, preserve as far as possible many from auto-component energy since the designed kernel function realized the possible match with the auto-component of FH signal, thus gained the better parameter estimation performance.The Cohen TF distributions are not only numerous, but also contains one or more parameters. The choice of kernel function type and parameter are important for the performance of TF distribution. It is major problem in TF analysis of FH signal to evaluate the performances of different distributions, select better type and parameters of TF distribution. In this thesis, we analysis the results of entropy measure in TF distribution performance evaluation and propose a novel optimization method about kernel parameters based on Renyi entropy with normalization volume. The parameter estimation variance is reduced effectively and estimation precision is increased based on Renyi entropy TF distribution, combining parameter optimization by sequential quadratic programming method.The TF analysis method is mainly divided into atom decomposition and the energy distribution according as the realization way. The former is to represent the signal as the weighted sun of TF elementary functions named atoms in order to eliminate the interference of cross-terms at all. In this thesis, we introduce the particle swarm optimization (PSO) algorithm to overcome the bottleneck of huge computation burden and design a novel parameter estimation algorithm of FH signal after obtaining atom parameters which match the FH signal components. This algorithm get rid of the Signal to Noise Ratio (SNR) threshold value effect of parameter estimation based on time-frequency plane and could obtain good estimate results under the low SNR condition.The PSO algorithm based on matching pursuit can only select one atom which matches the single component of FH signal in each iterate, thus the efficiency is reduced. In this thesis, we design a new PSO algorithm based on modified multi-species procedure which uses the algorithm convergence in the search procedure of the particles using multimodal characteractic of the inner product of FH signal and the atoms of the dictionary. Computer simulation results show that this algorithm could give all the TF atoms that match the components about FH signal and could enhance the parameter estimation performance under low SNR which is suitable for many kinds of FH signal.更多还原