摘要
人工混响是语音处理的重要研究方向。对此,介绍几种主流的人工混响算法。首先介绍全通滤波器串联模型,该模型结构最简单,但混响密度较低;其次介绍三种梳状滤波与全通滤波组合形式的混响模型,这类模型是目前比较常用的一类模型,能够达到较高的混响品质,并且运算量适中;再次,介绍基于反馈延时网络的混响模型,这类模型考虑各个延时线之间的交叉叠加,在平滑度、密度等方面均有进一步提升;最后介绍一种改进型的反馈延时网络混响模型,通过引入全通滤波器组和低通滤波器组增加了混响密度,优化了频率响应,能够达到较高的混响品质。
Artificial reverberation is an important study direction in speech processing.This paper introduces the characteristics of several mainstream artificial reverberation algorithms.Firstly,the series model of all-pass filter with simple structure is introduced,which has the simplest structure but low reverberation density.Then,three kinds of reverb models combined with comb filter and all-pass filter are introduced.These models are commonly used at present,which can achieve high reverb quality and moderate computation.Then,the reverberation model based on feedback delay network is introduced.This model takes into account the overlapping of various delay lines,which has further improved smoothness and density.Finally,an improved feedback delay network reverberation model is introduced.By introducing all-pass filter banks and low-pass filter banks,the reverberation density is increased,the frequency response is optimized,and the high reverberation quality can be achieved.
作者
沈平
邓小保
SHEN Ping;DENG Xiaobao(Changsha Echo Technology Co.,Ltd.,Changsha 710064,China)
出处
《电声技术》
2021年第10期9-14,共6页
Audio Engineering
关键词
人工混响
全通滤波器
梳状滤波器
反馈延时网络
artificial reverberation
all-pass filter
comb filter
feedback delay network
作者简介
沈平(1983-),男,博士,研究方向为系统建模与控制、音频信号处理。E-mail:ppguitar1983@sina.com;邓小保(1990-),男,硕士,研究方向为系统建模与控制、音频信号处理。
