基于改进EASI算法的多跳频信号盲源分离  

Multi-frequency-hopping Signals Blind Source Separation Based on Improved EASI Algorithm

作　　者：邓灵[1] 张天骐[1] 金静[1] 朱洪波[1] 

机构地区：[1]重庆邮电大学,信号与信息处理重庆市重点实验室,重庆400065

出　　处：《科学技术与工程》2014年第2期45-49,共5页Science Technology and Engineering

基　　金：国家自然科学基金项目(61071196,61102131);教育部新世纪优秀人才支持计划项目(NCET-10-0927);信号与信息处理重庆市市级重点实验室建设项目(CSTC2009CA2003);重庆市杰出青年基金项目(CSTC2011jjjq40002);重庆市自然科学基金项目(CSTC2010BB2398,CSTC2010BB2409,CSTC2010BB2411,CSTC2012JJA40008);重庆市教育委员会科研项目(KJ120525)资助

摘　　要：为了有效地解决多跳频信号的盲源分离问题,提出了一种变步长的EASI(Equivariant Adaptive Separation via Independence)盲源分离算法。该算法在传统EASI算法的等变化性基础上,用性能指标(串音误差)作为准则,通过改变函数的取值范围及形状,自适应更新步长,使其在一个固定小的范围内,达到算法收敛速度和稳定性能的一个较理想的平衡点,改善了当步长固定时存在的缺陷。经过实验仿真,证明该算法对步长有很好的调整能力,性能稳定且收敛速度较快,能很好地将多个跳频信号进行分离,较传统的EASI算法有更高的适用性。In order to effectively solve blind source separation of the multi-frequency-hopping signal, a variable step-size EASI algorithm based on blind source separation is proposed. The algorhhm use the Performance Index （crosstalk error） as a criterion, on the basis of equal variability of traditional EASI algorithm, to update the step adaptively by changing the range and shapes of the function. It can achieve a better balance between convergence speed and stability of the algorithm in a fixed small scope, so the algorithm has been improved compare to the fixed step. The experimental results show that the algorithm has a very good ability to adjust the step, performance is sta-ble and convergence speed is fast, multiple frequency hopping signals can be well separated, there is a higher availability compared to traditional EASI algorithm.

关 键 词：多跳频信号 盲源分离 EASI算法 变步长 

分 类 号：TN914.41[电子电信—通信与信息系统]