现代电子技术2024，Vol.47Issue(9)：71-77,7.DOI:10.16652/j.issn.1004-373x.2024.09.013
基于动量梯度下降的回声消除算法
Echo cancellation algorithm based on momentum gradient descent
陈张良 1卢敏 1曾桂根2
作者信息
- 1. 南京邮电大学 通信与信息工程学院,江苏 南京 210003
- 2. 南京邮电大学 通信与信息工程学院,江苏 南京 210003||南京邮电大学 通信与网络技术国家工程研究中心,江苏 南京 210003
- 折叠
摘要
Abstract
An improved proportional normalized least-mean-square algorithm based on the l0 norm(L0-IPNLMS)based on momentum gradient descent is proposed to address the issue of acoustic echoes affecting normal work in extreme environment speech systems and the slow convergence speed of conventional acoustic echo cancellation algorithms.In the algorithm,the momentum factor is introduced into the L0-IPNLMS algorithm to eliminate obvious gradient oscillation amplitude during algorithm operation,and improve the convergence speed of adaptive filters.In addition,the residual echo decreases significantly,which results in better acoustic echo suppression effect.Simulation experiments show that,in comparison with the L0-IPNLMS algorithm,the improved algorithm can reduce the mean-square error(MSE)by 3.47 dB and 3.69 dB and increase the echo return loss enhancement(ERLE)by 3.46 dB and 3.68 dB when simulating random multi-tone signals and real speech signal inputs.In low signal-to-noise ratio(SNR),the rate of convergence of the improved algorithm is higher than that of algorithms such as L0-IPNLMS,and its convergence effect is significantly improved when it is used for echo cancellation of real speech signals.关键词
回声消除算法/动量梯度下降/极端环境话音通信系统/归一化/最小均方算法/收敛速度Key words
echo cancellation algorithm/momentum gradient descent/extreme environment speech communication system/normalization/LMS algorithm/rate of convergence分类
信息技术与安全科学引用本文复制引用
陈张良,卢敏,曾桂根..基于动量梯度下降的回声消除算法[J].现代电子技术,2024,47(9):71-77,7.
