脉冲成形滤波器的优化设计  

Optimization design of pulse shaped filters

作　　者：薛强 胡迪 赵婕 张大元 

机构地区：[1]工业和信息化部电信研究院,北京100191

出　　处：《高技术通讯》2015年第10期971-979,共9页Chinese High Technology Letters

基　　金：国家重大科学仪器设备开发专项(2013YQ330667)资助项目

摘　　要：为了提高数字通信系统的信道带宽资源利用率和抑制码间干扰,研究了脉冲成形滤波器的优化设计,提出了采用神经网络算法的脉冲成形滤波器优化设计方法。首先分析了滚降因子、内插倍数及群时延等核心参数对滤波器性能的影响,获得了实际系统设计的依据,然后针对用传统窗函数法设计滤波器时难以精确控制过渡带宽的问题,提出了采用余弦基结构的神经网络算法。该算法可以实现对滤波器的边界频率的精确控制,取得较高的设计精度。研究表明,该算法通带范围无过冲、无波动,阻带衰耗大,初始条件可随机确定,是一种有效的设计方法。To effectively save digital communication systems＇ frequency resource and reduce their inter symbol interfer- ence, the optimization design of pulse shaped filters was studied, and an optimal design method using the neural network algorithm was proposed. Firstly, the influences of the key parameters of roll-off factor, interpolation multi- pie, and group delay effect on the filters＇ performance were analyzed to provide the basis for design of practical fil- ter systems, and then, aiming at the problem that the filter design using the traditional window function algorithm can not control the transition band of frequency response, a neural network algorithm using the cosine function- based structure was presented. This algorithm can realize the accurate control of the boundary frequency of pulse shaped filters to obtain the higher design accuracy. The proposed algorithm is proved effective in filter design by its characteristics of no overshoot or fluctuation in passband, and random determination of initial conditions and so on.

关 键 词：脉冲成形滤波器 MATLAB 窗函数法 神经网络法 

分 类 号：TN713[电子电信—电路与系统]