抑制多模噪声的变步长NLMS-Newton算法研究  

Research on NLMS-Newton algorithm with variable step length for multi-mode noise suppression

作　　者：祁霄 山拜.达拉拜 解倩倩 

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046 [2]新疆多语种信息技术重点实验室,新疆乌鲁木齐830046

出　　处：《现代电子技术》2018年第1期47-50,56,共5页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(60971130)~~

摘　　要：实际生活中产生的多模噪声会对信号产生严重的破坏甚至损失。传统LMS算法不仅收敛速度慢而且存在较大的稳态误差和失调量,故已不能很好地控制噪声。通过分析NLMS算法和Newton算法的优点,提出一种新的NLMS-Newton算法。该算法的收敛速度高,计算复杂度减小,且应用了随信噪比变化的变步长并修正了自相关矩阵,降低了算法收敛后的稳态误差和失调量。通过仿真表明,NLMS-Newton算法在收敛速度及稳态误差等方面都有了较大改善,并且能在很好地抑制多模噪声的同时提取出有用信号。Signal can be damaged and even lost because of the multi-mode noise generated in actual life. Since the traditional LMS algorithm has slow convergence speed, big steady-state error and high maladjustment, it can＇t control the noise perfectly. By analyzing the advantages of NLMS algorithm and Newton algorithm, a new NLMS-Newton algorithm is proposed. The algorithm has fast convergence speed and low calculating complexity, in which the step length changed with signal-to-noise ratio is applied and the autocorrelation matrix is corrected to reduce the steady-state error and maladjustment after the algorithm con- vergence. The simulation results show that the NLMS-Newton algorithm has improved greatly in the aspects of convergence speed and steady-state error, and can extract the useful signal while suppressing the multi-mode noise.

关 键 词：多模噪声抑制 NLMS算法 Newton算法 NLMS-Newton算法 自相关矩阵 变步长 

分 类 号：TN911-34[电子电信—通信与信息系统]