【作者】 王梓展；

【导师】 段正华；

【作者基本信息】 湖南大学 ， 通信与信息系统， 2005， 硕士

【摘要】 随着移动通信事业在全球的迅猛发展,人们对移动通信的容量和质量都有了越来越高的要求,提高频谱利用率刻不容缓,智能天线的发展给整个无线通信领域带来了新的生机。智能天线技术研究的核心是波束形成算法,找到一种真正快速收敛和性能优良的自适应算法是该研究的关键所在。其中,最小均方误差LMS(Least Mean Squares)算法,结构简单、易于实现,因而备受青睐。但是其固有的收敛速度和稳态失调之间的矛盾成为该算法广泛应用的瓶颈,本文就LMS算法展开研究,主要包括以下几个方面的工作: 1.比较了直线阵列和圆形阵列的优缺点,得出圆环阵更适用于移动通信系统的结论。重点对几种经典的智能天线波束形成算法进行分析,并分别指出其优劣,为LMS算法的研究提供理论基础。 2.针对变步长VSSLMS(Variable Step Size LMS)算法对噪声敏感问题,利用误差信号的相关值代替误差信号来调节步长,提出一种稳健的变步长自适应算法(Robust Variable Step-Size LMS:RVSSLMS)。理论分析和仿真结果表明该算法在不增加计算量的前提下不仅具有良好的抗干扰性,而且其收敛性能和稳态失调明显改善,同时算法对系统的突发变化也有良好的跟踪性能。 3.在分析变步长LMS算法性能的基础上,针对相关输入时,算法的收敛性能不佳的问题,借助解相关原理,用输入信号的正交分量来更新自适应滤波器的权系数,并对权值进行归一化处理,提出一种解相关变步长归一化LMS自适应算法(Uncorrelated Variable Step-Size Normalized LMS Algorithm:UVSSNLMS)有效提高了算法的收敛性能。理论分析和仿真实验表明改进算法对独立噪声有抗干扰性,归一化处理也扩大了算法步长的动态输入范围。 4.对提出的RVSSLMS和UVSSNLMS两种变步长算法,进行了对比分析、实验仿真,分别从抗干扰性、跟踪能力和计算复杂度几个方面进行比较,得出RVSSLMS算法更适合无线信道的结论。更多还原

【Abstract】 With the rapid development of mobile communication, its capacity and service quality is badly needed. To improve the frequency utilization becomes no time to delay. The technology of smart antenna brings great energy to the total wireless communication field. The core of smart antenna research is algorithm of beaming, whose rapid converging and excellent performance is very importance to implement the adaptive beaming. And the least mean squares (LMS) algorithm is often emphasized due to its simple structure and the convenience of realization. But its inherent contradiction between convergence rate and maladjustment of steady condition, confines its applications. In the paper, LMS algorithm is studied and the main work is as follows:1. The advantage and disadvantage of linear array and circular array are discussed respectively. Author analyze some classic beam-former algorithms and make comparison of them. Thereof, basic theory support of LMS algorithm is provided.2. A robust variable step size adaptive algorithm is proposed, who can effectively adjust the step size by utilizing the correlation characteristic of output signal with previous iteration error signal. The algorithm allows more flexible control of maladjustment and convergence time without the compromise one for the other and it can maintain the immunity against noise disturbance. The result of simulations also shows that the algorithm is robust to an abrupt change.3. Author analyze the variable step size LMS algorithm, and focus on the problem of reduction of performance of the modified variable step size LMS algorithm (MVSSLMS), due to the descent of its convergence speed, when the input signals heavy correlated. By the virtue of the principle of correlation, weight coefficients are updated by the orthogonal component of the input signal and normalized.As a result, an uncorrelated variable step size normalized LMS algorithm (UVSSNLMS) is proposed, which can achieve better performance. The theoretical analysis and result of simulations supports that the modified algorithm is immunized from noise and the process of norm extends the dynamic input of step size.4. The two newly proposed algorithms, RVSSLMS and UVSSNLMS,are analyzed and simulated in some aspects, such as disturbance resistance, tracking capability and complexity. In conclusion, the RVSSLMS is more suitable for wireless更多还原