多窗谱小波阈值估计的维纳滤波语音增强算法
Wiener filter speech enhancement algorithm base on multi-taper spectral wavelet threshold estimation
罗芳 1李志亮1
作者信息
- 1. 宁德师范学院 信息工程学院,福建 宁德 352100
- 折叠
摘要
提出一种多窗谱小波阈值估计的维纳滤波语音增强算法.该算法采用多窗谱估计噪声功率谱,结合改进的小波阈值法减小其误差,再计算先验信噪比.改进后的阈值函数具有连续性,增加了小波系数的平滑度,实现了信噪比的平滑.与传统软、硬阈值函数算法相比,文中算法具有更高的信噪比,能较好地抑制噪声,有效提高语音清晰度.
Abstract
The Wiener filtering algorithm with the multi window spectral wavelet threshold estimation method is proposed.A Wiener filtering speech enhancement algorithm is combined with multi window spectral wavelet threshold estimation.The algorithm adopts a multi window spectrum to estimate the noise power spectrum,com-bined with an improved wavelet threshold method to reduce its error,and then calculates the prior signal-to-noise ratio.The improved threshold function has continuity,increases the smoothness of wavelet coefficients,and achieves smooth signal-to-noise ratio.The experimental results show that compared with traditional soft and hard threshold function algorithms,the improved algorithm has a higher signal-to-noise ratio,suppresses the residual noise in the stable noise environment,and improves the speech intelligibility.
关键词
语音增强/维纳滤波/小波阈值/多窗谱Key words
speech enhancement/Wiener filtering/wavelet threshold/multi-taper spectrum引用本文复制引用
基金项目
福建省自然科学基金项目(2022J011224)
福建省自然科学基金项目(2021J011169)
福建省中青年教师教育科研项目(JAT210464)
宁德师范学院中青年科研项目(2020Q107)
出版年
2024
NETL
NSTL
万方数据