数字助听器中声反馈消除算法的研究

发布时间：2018-05-14 12:24

本文选题：数字助听器 + 声反馈消除　；参考：《南京邮电大学》2015年硕士论文

【摘要】：佩戴数字助听器可以有效改善听损患者的听力水平,然而数字助听器中扬声器和麦克风位置很接近,容易产生声反馈现象,严重影响助听器的性能。自适应声反馈消除算法可以有效地消除声反馈信号,提高助听器输出语音质量。本文重点研究助听器中的声反馈消除算法,并提出改进算法。论文的主要工作如下:(1)研究了语音的发音机理及其特性,依次讨论了激励模型、声道模型和辐射模型这三个语音信号产生模型的主要组成部分。依次介绍了时域分析、频域分析和倒谱分析这三种常用的语音信号分析方法。分别介绍了声反馈消除、响度补偿与压缩频移、语音增强和声源定位这几种数字助听器中的关键技术。比较了LMS、归一化LMS(NLMS)和RLS这三种常用的自适应滤波算法,并通过仿真实验对比了三种算法的跟踪性能。(2)固定步长Improved Propotion NLMS(IPNLMS)算法应用于声反馈消除时,必须在收敛速度和稳态失调折中取一个合适的步长因子。为了解决这一对矛盾,本文提出一种新的变步长方法代替固定步长方法。与固定步长和传统的变步长IPNLMS算法相比,该算法收敛速度更快,稳态失调更小,声反馈消除效果更加明显。(3)传统的NLMS和Propotion NLMS(PNLMS)算法在处理语音信号时,性能有所降低。为了对数字助听器中声反馈信号进行有效消除,本文采用APA算法代替传统的声反馈消除算法对助听器中声反馈信号进行消除。为了提高算法的实用性,本文提出一种简化的、变步长的APA算法代替固定步长APA算法。与固定步长APA算法相比,该算法计算量进一步减小,并且收敛速度和稳态失调性能都得到进一步提高。
[Abstract]:Wearing digital hearing aids can effectively improve the hearing level of patients with hearing impairment. However, the loudspeakers and microphones in digital hearing aids are very close to each other, which is easy to produce acoustic feedback, which seriously affects the performance of hearing aids. Adaptive acoustic feedback cancellation algorithm can effectively eliminate acoustic feedback signals and improve the quality of hearing aid output speech. This paper focuses on the acoustic feedback elimination algorithm in hearing aids, and proposes an improved algorithm. The main work of this paper is as follows: (1) the phonetic mechanism and characteristics of speech are studied, and the three main components of speech signal generation model, excitation model, vocal channel model and radiation model, are discussed in turn. Time domain analysis, frequency domain analysis and cepstrum analysis are introduced in turn. The key technologies of digital hearing aid, such as acoustic feedback cancellation, loudness compensation and compression frequency shift, speech enhancement and sound source location, are introduced respectively. Three commonly used adaptive filtering algorithms, LMS, normalized LMSNLMS) and RLS, are compared. The tracking performance of the three algorithms is compared by simulation experiments. The fixed-step Improved Propotion NLMS / IPNLMS) algorithm is applied to acoustic feedback cancellation. An appropriate step factor must be taken between the convergence rate and the steady state misalignment. In order to solve this contradiction, a new variable step size method is proposed to replace the fixed step size method. Compared with the fixed step size and the traditional variable step size IPNLMS algorithm, the proposed algorithm has faster convergence speed, less steady-state misalignment, and more obvious acoustical feedback cancellation effect.) the traditional NLMS and Propotion NLMS PNLMS) algorithms have lower performance in processing speech signals. In order to effectively eliminate the acoustic feedback signal in digital hearing aid, APA algorithm is used instead of the traditional acoustic feedback elimination algorithm to eliminate the acoustic feedback signal in the hearing aid. In order to improve the practicability of the algorithm, a simplified, variable step size APA algorithm is proposed to replace the fixed step size APA algorithm. Compared with the fixed-step APA algorithm, the computational complexity of the algorithm is further reduced, and the convergence rate and steady-state misalignment performance are further improved.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN912.3;R764.5

【相似文献】