【作者】 池越；

【导师】 陈国鹰；

【作者基本信息】 河北工业大学 ， 微电子学与固体电子学， 2011， 博士

【摘要】 偏振模色散是限制光纤通信系统的传输速率和传输距离进一步提高的主要因素,因此,对于偏振模色散补偿技术的研究具有重要的理论价值和应用价值。本论文主要利用改进的粒子群算法和支持向量机等计算智能方法解决偏振模色散自适应补偿中的一些关键问题,主要研究内容和创新工作如下:(1)针对基本粒子群优化算法(BPSO)存在易于陷入局部极值、粒子速度有限等缺陷,结合量子理论,提出了具有量子行为的QPSO算法,通过与遗传算法、BPSO算法、具有梯度搜索因子的改进PSO算法等算法的比较发现:QPSO算法具有收敛速度快、不易陷入局部极值等优点。在此基础上,还针对QPSO算法在处理离散数据方面的不足,提出一种新的QDPSO算法,为优化求解离散问题提供了新的思路。(2)在PMD电域自适应均衡系统中,自适应算法的作用至关重要。由于常见的自适应算法(如LMS算法、RLS算法等)存在收敛速度慢、运算量巨大等缺馅,从而影响了电域自适应均衡器的均衡效果。针对这一问题,提出基于QPSO自适应算法的PMD电域自适应均衡方案,获得较常见的自适应算法更好的均衡效果,解决了由均衡器自适应算法性能缺陷带来的均衡效果不佳的问题。(3)在PMD光域自适应补偿系统中,控制算法是自适应补偿单元的核心。由于常见的控制算法(如单纯形算法、GA算法等)存在收敛速度较慢、容易陷入局部极值等缺馅,从而影响了光域自适应补偿器的补偿效果。针对这一问题,提出基于QPSO控制算法的PMD光域自适应补偿方案,获得较其它常见的控制算法更好的补偿效果,解决了由补偿器控制算法性能缺陷带来的补偿效果不佳的问题。(4)提出一种带有信号调制格式识别功能的PMD自适应补偿方案。未来的全光网络必然是多种调制格式的混合传输网,为此在对带有PMD的光信号进行补偿之前,加入一个信号调制格式识别环节,即利用支持向量机来准确识别信号不同的调制格式,然后根据识别结果,选用对该调制格式补偿效果最好的补偿方案对其进行补偿。(5)基于支持向量机技术,设计并实现了调制信号识别方案。完成了幅度调制(AM)、频率调制(FM)和相位调制(PM)等三类调制信号的识别,最终获得了94.4%的识别准确率,能很好满足实际需求。更多还原

【Abstract】 Polarization Mode Dispersion (PMD) is the main factor of limiting the further improve for the transmission rate and transmission distance of the optical fiber communication system. It has the important theoretical value and application value to take the PMD compensation technology study.This paper tries to solve the key problems in the PMD adaptive compensation using the computational intelligence method, such as improved Particle Swarm Optimization (PSO) and Support Vector Machines (SVM). Main research contents and innovation work are as follows:(1)Since the basic Particle Swarm Optimization (BPSO) is trapped in local optimum and has finite particle velocity, this paper put forward QPSO algorithm possesses the uantum behaviors combined with quantum theory. Compared with the genetic algorithm, BPSO algorithm and the improved PSO algorithm with gradient search factor,QPSO algorithm has fast convergence, and is not easy to be trapped in local optimum. On this basis, the paper takes out the new QDPSO algorithm to give the new idea for optimizing and solving the discrete problem, according to the deficiency of the QPSO algorithm on the discrete data treatment.(2)It plays the most important role of the adaptive algorithm in the PMD optical domain adaptive equalizing system. Equalization result of the optical domain adaptive equalizer has been influenced, because of the slow convergence speed and huge operation of the normal adaptive algorithm(such as LMS algorithm, RLS algorithm).On this basis, the PMD optical domain adaptive equalizing scheme based on QPSO adaptive algorithm has been first proposed,which compared with the normal adaptive algorithm, the better equalization result has been obtained and the poor equalization result problem which brings by the limitation of the properties for the equalizer adaptive algorithm has been solved.(3)Control algorithm is the core of the adaptive compensation unit in the PMD optical domain adaptive compensation system. Compensation effect of the optical domain adaptive compensator has been influenced, because of the slow convergence speed and be easily trapped in local optimum of the normal control algorithm (such as simplex algorithm, GA algorithm). On this basis, the PMD optical domain adaptive compensation scheme based on QPSO adaptive algorithm has been first proposed,which compared with the normal control algorithm, the better equalization result has been obtained and the poor compensation effect problem which brings by the limitation of the properties for the compensator control algorithm has been solved. (4) The PMD adaptive compensation scheme with the recognition function of signal modulation mode has been first put forwarded. The differ modulation signals has been transmitted in the future all optical network. The signal modulation mode recognition link which use the Support Vector Machines to accurately identify the different signal modulation mode has been added before the compensation of the optical signal with PMD. Then according to the identification result, the compensation has been made by using the compensation scheme which has the best compensation effect for the modulation mode, in order to get the further improvement of the compensation effect.(5) The signal modulation recognition scheme has been designed and realized in this paper,which based on the Support Vector Machines (SVM). The Amplitude modulation, Frequency modulation,and Phase modulation have been done to get the recognition accuracy of 94.4%, which can meet the actual demand very commendably.更多还原