【摘要】 对DCT域基于拉普拉斯统计模型的语音增强,分析了模型因子的估计误差及其对于算法整体增强性能的影响,并根据广义高斯分布模型及其形态参数的概念与性质,提出了一种新的拉普拉斯模型因子估计方法,该方法结构简单,它利用拉普拉斯模型条件下语音分量方差与模型因子的对应关系,间接地获取模型因子的估计,算法不仅有效地消除了噪声分量对于估计精度的影响,而且可以快速地跟踪语音分量的变化。仿真结果表明,基于该模型因子估计方法的语音增强算法在多种噪声背景下具有更出色的语音增强效果。更多还原

【Abstract】 Aimed at the enhancement of noisy speech using the Laplacian model in a discrete cosine transform domain,a novel approach for Laplacian factor estimation is presented based on the property of generalized Gaussian distribution model and its shape parameter.The proposed approach can indirectly obtain the estimation of the Laplacian factor using its relation with the variance of clean speech components under the Laplacian distribution assumption,thus the method is simple.The algorithm can eliminate the effect by noise components and give an accurate estimation for the Laplacian factor.Furthermore,it can fast track the changes of speech components with one frame delay.Experimental results demonstrate the improved performance of the proposed algorithm in different noises.更多还原