基于步长最优化的EASI盲源分离算法
Step-size Optimization Based EASI Algorithm for Blind Source Separation
作者：付卫红(江南电子通信研究所 通信系统信息控制技术国家级重点实验室，浙江 嘉兴 314001)；杨小牛(江南电子通信研究所 通信系统信息控制技术国家级重点实验室，浙江 嘉兴 314001)；刘乃安(西安电子科技大学 ISN国家重点实验室，陕西 西安 710071)
Author：(National lab. of Info. Control Technol. for Communication System, Jiangnan Electronic Communications Research Inst., Jiaxing 314033, China)；(National lab. of Info. Control Technol. for Communication System, Jiangnan Electronic Communications Research Inst., Jiaxing 314033, China)；(State Key Lab. of Integrated Service Networks, Xidian Univ., Xi’an 710071, China)
收稿日期：2006-10-17 年卷（期）页码：2008,40(1):118-121
期刊名称：工程科学与技术
Journal Name：Advanced Engineering Sciences
关键字：盲源分离;通信信号处理;步长最优化;收敛速度
Key words：blind source separation；communication signal processing；step-size optimization；convergence speed
基金项目：国家自然科学基金资助项目(60672038;60702060)
中文摘要
针对通信信号的盲分离问题，在EASI批处理算法的基础上，提出了一种基于步长最优化的EASI盲源分离算法（SO EASI）。该算法基于最速下降法的思想来调整步长，因此该算法不论迭代步长的初始值如何选取，其收敛速度相对固定步长的迭代算法都有很大幅度的提高，计算机仿真验证了该算法对通信信号分离的有效性。结果表明当初始步长为0.02时，其算法的收敛速度比固定步长的EASI算法以及VS EASI算法的收敛速度提高10倍以上。
英文摘要
Aim at the problem of blind source separation of communication signals, a step size optimization equivariant adaptive source separation via independence (SO-EASI) algorithm was proposed based on the EASI algorithm. This algorithm adjusted the step size based on the steepest descent method and thereby greatly improved its convergence speed whether the step-size was initialized at a bigger or smaller positive number. Simulation results proved that SO EASI algorithm can effectively separate the communication signals blindly. Results showed that when the initial step size is equal to 0.02, the convergence speed of the SO-EAIS algorithm is 10 times that of fixed-step EASI algorithm and VS-EASI algorithm.
【关闭】
