具有约束学习速率的频域盲均衡算法  

Blind Equalization Algorithm in Frequency Domain Characterised with Restricted Learning Rate

作　　者：沈希忠[1] 许海翔[1] 史习智[1] 

机构地区：[1]上海交通大学振动冲击噪声国家重点实验室

出　　处：《电子学报》2005年第4期762-765,共4页Acta Electronica Sinica

基　　金：国家自然科学(No.60372075);上海市教委资助项目(No.03YQHB160);上海市科委资助项目(No.04ZR14126)

摘　　要：多通道的盲均衡在语音分离、去混响、通信、信号处理和控制等领域具有广泛的应用,本文基于二阶统计量在频域重新构造评价函数,为避免得到平凡解在评价函数中额外引入了一项不可简约多项式的描述.然后,应用自然梯度法导出盲均衡算法,为了确保算法收敛,同时给出迭代收敛条件,并将它用来约束学习速率.仿真实验表明,该算法能够分离可均衡的有限冲激响应系统,从而证实了该算法的有效性.Multichannel blind equalization is an important task for numerous applications such as speech separation, dereverberation, communication, signal processing and control, etc. We reconstruct a cost function based on second order statistics. To avoid trivial solution, an additional item is added to the cost function, which assures that the equalizer is irreducible. Then, an algorithm is derived with natural gradient search method for multichannel blind source separation and deconvolution of convolved signal mixtures. To avoid divergence of the algorithm, an iterative converging condition is also given, and is used to adjust the learning rate. Simulations indicate the ability of the algorithm to perform blind equalization under the weaker condition (the FIR system is equal).

关 键 词：盲均衡 评价函数 自然梯度法 学习速率 

分 类 号：TN911.7[电子电信—通信与信息系统]