基于改进鲸鱼算法的数字助听器回声消除方法  

Echo Cancellation Method of Digital Hearing Aids Based on Improved Whale Algorithm

作　　者：熊书慧 胡开明[1] 吴光文[1,2] 李跃忠 Xiong Shuhui;Hu Kaiming;Wu Guangwen;Li Yuezhong(School of Mechanical and Electronic Engineering,East China University of Technology,Nanchang 330013,China;Jiangxi Rehabilitation Auxiliary Industry Technology Research Institute,Nanchang 330013,China)

机构地区：[1]东华理工大学机械与电子工程学院,南昌330013 [2]江西省康复辅具产业技术研究院,南昌330013

出　　处：《机电工程技术》2024年第4期147-151,共5页Mechanical & Electrical Engineering Technology

基　　金：国家自然科学基金项目(61663001);江西省科技厅重点研发计划项目(20212BBE53033)。

摘　　要：在数字助听器应用中,回声消除能力是检验助听器效果的标准之一。针对回声消除算法中自适应滤波器权重设置主观性较强以及实际语音信号波动性较大导致数字助听器回声消除精度不高的问题,提出了一种结合集合经验模态分解(EEMD)方法及鲸鱼算法(WOA)的最小均方误差算法(LMS)。该新方法在鲸鱼算法的基础上引入集合经验模态分解方法,采用集合经验模态分解方法将采集到的语音信号分解为若干个平稳的子序列,运用鲸鱼算法收敛速度较快的特点帮助自适应滤波器更快找到更适宜的权重,对自适应滤波器进行权重向量的优化,从而建立EEMD-WOA-LMS模型,通过回声消除效果的常用性能指标对所提方法进行验证。仿真实验表明,将改进方法与WOA-LMS算法相比,能获得较快的收敛速度和较低的失调量,并且均方误差有所降低,失调量降低了0.73 dB,回声损失值提升了2.87 dB,验证了EEMD-WOA-LMS算法的回声消除能力更优。In digital hearing aid applications,echo elimination ability is one of the standards for testing the effects of hearing aids.For the echo elimination algorithm,the adaptive filter weight is set with strong subjectivity and the actual voice signal volatility,resulting in the problem of the low accuracy of digital hearing aid echo,a minimum mean square error(LMS)algorithm combining set empirical mode decomposition(EEMD)method and whale algorithm(WOA)is proposed.This new method introduces the set empirical mode decomposition method based on the whale algorithm,by utilizing the fast convergence speed of the whale algorithm,the adaptive filter can quickly find more suitable weights,optimize the weight vector of the adaptive filter,and establish an EEMD-WOA-LMS model.The proposed method is validated through commonly used performance indicators of echo cancellation effect.The simulation experiments show that compared with the WOA-LMS algorithm,the improvement method can obtain a faster convergence speed and lower imbalance,and the all directions are reduced.The misalignment decreased by 0.73 dB and the echo loss value increased by 2.87 dB,it is verified that the EEMD-WOA-LMS algorithm has better echo elimination capacity.

关 键 词：集合经验模态分解 自适应滤波器 鲸鱼算法 数字助听器 回声消除 

分 类 号：TN912.3[电子电信—通信与信息系统]