摘要 通过讨论纯净语音分量的概率分布特征以及相邻分量间的统计相关特性,在自适应K-L变换(KLT,Karhunen-Loève Transform)域给出了一种新的语音信号统计模型,然后基于该信号模型,利用最大后验(MAP,Maximum a Posterior)估计理论提出了一种新型的单通道语音增强算法.该算法充分考虑到在KLT域相邻时刻语音分量间存在的相关信息,利用信号的高斯模型假设条件,以联合概率密度函数的形式将这种相关信息融合到MAP中,获得纯净语音分量的估计.算法不仅结构简单利于实现,且有效地避免了传统算法对语音分量估计的不足.仿真结果表明本文算法在客观和主观测试中都具有较好的语音增强效果.
Abstract：This paper investigates the probability distribution of speech components as well as the correlation characteristic between adjacent components,and presents a new speech model for enhancing noisy speech in adaptive KLT domain.Based on this model,a novel speech enhancement algorithm using MAP estimation is proposed,which incorporates the inter-frame correlation information as a form of joint probability density function into MAP under Gaussian model assumption for speech and noise components.The obtained estimation result keeps simple and avoids deficiency of classic approaches in enhancing noisy speech.In simulations with speech signals degraded by various noises,the proposed algorithm shows improved performance for a number of objective and subjective measures.
欧世峰;赵晓晖. 基于帧间相关性的最大后验估计语音增强算法[J]. 电子学报, 2007, 35(10): 2007-2013.
OU Shi-feng;ZHAO Xiao-hui. MAP Estimation for Noisy Speech Enhancement Based on Inter-Frame Correlation. Chinese Journal of Electronics, 2007, 35(10): 2007-2013.
2007, Vol. 35 
