用后验信噪比修正小波包自适应阈值的语音增强算法  

Speech enhancement algorithm based on wavelet packet adaptive threshold revised by posteriori SNR

作　　者：张雪英[1] 任永梅[1] 贾海蓉[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《中南大学学报（自然科学版）》2013年第11期4566-4573,共8页Journal of Central South University:Science and Technology

基　　金：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)

摘　　要：针对固定阈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阈值的语音增强算法。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阈值进行平滑,保证了阈值的连续性;进一步用指数化的后验信噪比自适应修正阈值,减少语音失真。实验结果表明:新算法提高了增强语音的信噪比和分段信噪比,与固定阈值小波包语音增强算法相比,具有更好的增强效果。Aiming at the problem that fixed threshold wavelet packet speech enhancement algorithm causes the speech distortion and residual noise, the 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 frame wavelet packet threshold which changed with the scale was smoothed by the sigmoid function with the parameter of posteriori SNR, which ensured the continuity of threshold. Furthermore, this threshold was adaptively revised further by the exponential posterior SNR, which reduced speech distortion. The experimental results show that the proposed algorithm improves the SN-R and segment SNR （SSNR） of the enhanced speech; compared with the fixed threshold wavelet packet speech enhancement algorithm, the proposed algorithm has better enhanced effect.

关 键 词：小波包 自适应阈值 噪声估计 后验信噪比 语音增强 

分 类 号：TN912.3[电子电信—通信与信息系统]