混响背景下基于分数阶傅里叶变换的自适应LMS滤波算法  

Adaptive LMS filtering algorithm based on fractional Fourier transform in reverberation background

作　　者：马凯 王易川[1] 陈喆 程玉胜[1] MA Kai;WANG Yichuan;CHEN Zhe;CHENG Yusheng(Navigation and Observation Department,Navy Submarine Academy,Qingdao 266000,Shandong,China)

机构地区：[1]海军潜艇学院航海观通系,山东青岛266000

出　　处：《声学技术》2020年第6期769-773,共5页Technical Acoustics

摘　　要：针对强混响背景下经典的最小均方误差(Least Mean Square,LMS)滤波算法难以有效地实现信混分离的问题,提出一种基于分数阶傅里叶变换的自适应LMS算法。首先将混响信号和自适应LMS滤波算法中的参考信号进行分数阶傅里叶变换,寻找最优变换域,并在分数阶域进行带通滤波,然后将得到的信号进行分数阶傅里叶反变换,最后将基于正态分布曲线的变步长LMS算法应用于此混响条件下进行滤波。仿真和海试数据验证结果表明,在信混比为0 dB的情况下,算法仍可以有效地滤除混响,使信混比提高6 dB。Aiming at the problem that the classical least mean square(LMS)filtering algorithm in the strong reverberation background is difficult to separate signal and reverberation effectively,an adaptive LMS algorithm based on fractional Fourier transform(FRFT)is proposed.Firstly,FRFT is performed to the reverberation signal and the reference signal of adaptive LMS filtering algorithm to find out the optimal transform domain,and the band-pass filtering is performed in the fractional order domain,then the fractional Fourier inverse transform(FRFIT)is performed to the obtained signal;and finally the variable-step-size LMS algorithm based on normal distribution curve is used for filtering in this reverberation condition.The verification results of simulation and sea-trial show that the algorithm can effectively filter out the reverberation under a signal-to-reverberation ratio of 0.3 dB,and the signal-to-reverberation ratio is improved by 6 dB.

关 键 词：混响 分数阶傅里叶变换 自适应LMS滤波 变步长 

分 类 号：TN911.7[电子电信—通信与信息系统]