摘要
为了克服低信噪比输入下,语音增强造成语音清音中的弱分量损失,造成重构信号包络失真的问题。论文提出了一种新的语音增强方法。该方法根据语音感知模型,采用不完全小波包分解拟合语音临界频带,并对语音按子带能量进行清浊音区分处理,在阈值计算上,提出了一种清浊音分离,基于子带信号能量的小波包自适应阈值算法。通过仿真实验,客观评测和听音测试表明,该算法在低信噪比输入时较传统算法,能够更加有效地减少重构信号包络失真,在不损伤语音清晰度和自然度的前提下,使输出信噪比明显提高。将该算法与能量谱减法结合,进行二次增强能进一步提高降噪输出的语音质量。
When input signal has low SNR,the commonly used wavelet pocket de-noising algorithm will cause envelope distortion problem for reconstructed signal because of unvoiced information losses.In order to overcome this,this paper presents a new method for speech enhancement.Motivated by speech perception model,incomplete wavelet packet decomposition are used to fit speech critical band,and the voiced and unvoiced sounds are processed separately based on sub-band energy ratio.A new wavelet threshold algorithm is obtained based on sub-band signal to noise energy ratio.In our comparative simulation test,results of objective evaluation and subjective test show that the proposed algorithm is more effective than traditional algorithm for low signal-to-noise ratio input,and it either removes noise as much as possible to improve the output SNR,or effectively reduces signal reconstruction distortion without doing harm to clarity and naturalness of speech intelligibility premise.When this new algorithm is combined with energy spectral subtraction,it can further improve the quality of speech de-noising.
出处
《应用声学》
CSCD
北大核心
2011年第1期72-80,共9页
Journal of Applied Acoustics
关键词
语音降噪
小波包分解
自适应阈值算法
子带能量
Wavelet packet threshold
Hearing masking
Speech enhancement
Adaptive algorithm