适用于助听器语音信号处理的DFT调制滤波器组设计  

Design of DFT modulated filter bank for speech processing in hearing aids

作　　者：蔡宇[1,2] 侯朝焕[1] 原建平[1] 鄢社锋[1] 

机构地区：[1]中国科学院声学研究所,北京100190 [2]中国科学院研究生院,北京100049

出　　处：《声学技术》2012年第5期517-521,共5页Technical Acoustics

基　　金：国家自然科学基金资助项目(60802072)

摘　　要：提出了一种近似重构、过采样DFT调制滤波器组的设计方法。根据滤波器组的系统失真和输出混迭方面的要求,构造一个简单的目标函数,并采用等纹波法优化设计原型滤波器。与现有方法相比,该设计方法将复杂的非线性优化问题转化为对单一参数的选取,简化了设计过程,并且可在满足DFT调制滤波器组设计准则的条件下使得通带范围内误差逼近最小。利用加权叠接相加(Weighted Overlap and Add,WOLA)的快速实现结构,使整个系统更加高效、易行,可用于助听器语音信号分频带处理与综合。This paper proposes a simple method of designing an oversampled,near-perfect reconstruction DFT modulated filter bank.According to the requirement of system distortion and aliasing power,a simple objective function is introduced,and the prototype filter is designed with the equiripple method.Compared with traditional methods,the proposed method optimizes a single parameter instead of the complex nonlinear optimization,and the design process is simplified.This approach can minimize the passband error under the filter-bank design criteria.By using weighted overlap and add（WOLA） structure,the system can be implemented efficiently,which is suitable for speech signal analysis and synthesis in hearing aids.

关 键 词：DFT调制滤波器组 等纹波法 WOLA结构 语音信号处理 

分 类 号：TN713[电子电信—电路与系统]