第
44
卷第
11
期
2013
年
11
月
中南大学学报(自然科学版)
Journal
of
Central
South University (Science
and
Technology)
Vol.44 No.11
Nov.
2013
用后验信。呆比修正小波包自适应阁值的语音增强算法
张雪英,任永悔,贾海蓉
(太原理工大学信息工程学院,山西太原,
030024)
摘要:针对固定阂值小波包语音增强算法造成的语音失真问题,提出一种采用后验信噪比修正小波包自适应阂值
的语音增强算法。该算法先用结合掩蔽效应改进的非平稳噪声估计算法估计噪声功率,确保计算出准确的节点后
验信噪比;再用含有此后验信噪比的
Sigmoid
函数对相邻帧的随尺度变化的阂值进行平滑,保证了阂值的连续’性;
进一步用指数化的后验信噪比自适应修正阂值,减少语音失真。实验结呆表明:新算法提高了增强语音的信噪比
和分段信噪比,与固定阂值小波包语音增强算法相比,具有更好的增强效呆。
关键词:小波包;自适应阂值;噪声估计;后验信噪比;语音增强
中图分类号:
TN912.3
文献标志码:
A
文章编号:
1672-7207(2013)11-4566-08
Speech enhancement algorithm based on wavelet packet adaptive
threshold revised
by
posteriori
SNR
ZHANG
Xueying,
REN
Yongmei
,丑
A
Hairong
(College
of
Information
Engin
巳
d
吨,
Tai
严
mn
University
of
Technology, Taiyuan 030024, China)
Abstract: Aiming at the problem that fixed threshold wavelet packet speech enhancement algorithm causes the speech
distortion and residual
noise
,出巳
speech
enhancement algorithm which used the posteriori signal-to-noise ratio (SNR)
of
wavelet packet nodes to adaptively revise threshold was proposed.
In
this algorithm, the noise power was estimated by
the improved non-stationary noise estimation algorithm that combined with masking effect, which ensured that posterior
SNR
of
nodes was calculated accurately. The
adjacent
企缸
m
wavelet packet threshold which changed with the scale was
smoothed by the
sigmoid
如
nction
with the
p
缸缸
neter
of
posteriori SNR, which
ensured
出巳
continuity
of
threshold.
Fur
出
ermore,
this threshold was adaptively
revised
如
rther
by
th
巳巳
xponential
posterior SNR, which reduced speech
distortion. The experimental results show that
th巳
proposed
algorithm improves the SNR and segment SNR (SSNR)
of
the enhanced speech;
comp
缸巳
d
wi
出
the
fixed threshold wavelet packet speech enhancement algorithm,
th巳
proposed
algorithm has better enhanced effect.
Key words: wavelet packet; adaptive threshold; noise estimation; posterior SNR; speech
ei
由
an
cement
语音增强的应用广泛,因此寻找到一种有效算法
来滤除带噪语音信号中的噪声,达到较强的抗噪效果
是十分重要的[!]。语音增强的目的就是要去除掉背景
噪声,提高语音的质量和可懂度,使昕音人在乐于接
收稿日期
z
2012
一
10
一
18
:修回日期
z
2012
一
12-22
受增强语音的同时能够更好的理解语音,而这些目的
通常不能兼得,要根据具体应用有所侧重。小波分析
理论是一种运用多分辨率思想分析非平稳信号的方
法。到目前为止,小波阙值语音增强方法由于算法简
基金项曰:国家自然科学基金资助项目(
61072087'
61370093
):山西省留学回国人员科研基金资助项目(
2011-035);
2012
年度山西省高校高新技术
产业化项目(
20120010)
通信作者
z
贾海蓉(
1977
一),女,山西洪
i
同人,博士,副教授,从事语音信号处理、
DSP
应用、多媒体通信等研究;电话
z
15536925678; E-mail
he
le
呵
ia722@163.com
第
44
卷第
11
期
2013
年
11
月
中南大学学报(自然科学版)
Journal
of
Central
South University (Science
and
Technology)
Vol.44 No.11
Nov.
2013
用后验信。呆比修正小波包自适应阁值的语音增强算法
张雪英,任永悔,贾海蓉
(太原理工大学信息工程学院,山西太原,
030024)
摘要:针对固定阂值小波包语音增强算法造成的语音失真问题,提出一种采用后验信噪比修正小波包自适应阂值
的语音增强算法。该算法先用结合掩蔽效应改进的非平稳噪声估计算法估计噪声功率,确保计算出准确的节点后
验信噪比;再用含有此后验信噪比的
Sigmoid
函数对相邻帧的随尺度变化的阂值进行平滑,保证了阂值的连续’性;
进一步用指数化的后验信噪比自适应修正阂值,减少语音失真。实验结呆表明:新算法提高了增强语音的信噪比
和分段信噪比,与固定阂值小波包语音增强算法相比,具有更好的增强效呆。
关键词:小波包;自适应阂值;噪声估计;后验信噪比;语音增强
中图分类号:
TN912.3
文献标志码:
A
文章编号:
1672-7207(2013)11-4566-08
Speech enhancement algorithm based on wavelet packet adaptive
threshold revised
by
posteriori
SNR
ZHANG
Xueying,
REN
Yongmei
,丑
A
Hairong
(College
of
Information
Engin
巳
d
吨,
Tai
严
mn
University
of
Technology, Taiyuan 030024, China)
Abstract: Aiming at the problem that fixed threshold wavelet packet speech enhancement algorithm causes the speech
distortion and residual
noise
,出巳
speech
enhancement algorithm which used the posteriori signal-to-noise ratio (SNR)
of
wavelet packet nodes to adaptively revise threshold was proposed.
In
this algorithm, the noise power was estimated by
the improved non-stationary noise estimation algorithm that combined with masking effect, which ensured that posterior
SNR
of
nodes was calculated accurately. The
adjacent
企缸
m
wavelet packet threshold which changed with the scale was
smoothed by the
sigmoid
如
nction
with the
p
缸缸
neter
of
posteriori SNR, which
ensured
出巳
continuity
of
threshold.
Fur
出
ermore,
this threshold was adaptively
revised
如
rther
by
th
巳巳
xponential
posterior SNR, which reduced speech
distortion. The experimental results show that
th巳
proposed
algorithm improves the SNR and segment SNR (SSNR)
of
the enhanced speech;
comp
缸巳
d
wi
出
the
fixed threshold wavelet packet speech enhancement algorithm,
th巳
proposed
algorithm has better enhanced effect.
Key words: wavelet packet; adaptive threshold; noise estimation; posterior SNR; speech
ei
由
an
cement
语音增强的应用广泛,因此寻找到一种有效算法
来滤除带噪语音信号中的噪声,达到较强的抗噪效果
是十分重要的[!]。语音增强的目的就是要去除掉背景
噪声,提高语音的质量和可懂度,使昕音人在乐于接
收稿日期
z
2012
一
10
一
18
:修回日期
z
2012
一
12-22
受增强语音的同时能够更好的理解语音,而这些目的
通常不能兼得,要根据具体应用有所侧重。小波分析
理论是一种运用多分辨率思想分析非平稳信号的方
法。到目前为止,小波阙值语音增强方法由于算法简
基金项曰:国家自然科学基金资助项目(
61072087'
61370093
):山西省留学回国人员科研基金资助项目(
2011-035);
2012
年度山西省高校高新技术
产业化项目(
20120010)
通信作者
z
贾海蓉(
1977
一),女,山西洪
i
同人,博士,副教授,从事语音信号处理、
DSP
应用、多媒体通信等研究;电话
z
15536925678; E-mail
he
le
呵
ia722@163.com