摘要
针对非平稳环境噪声提出一种基于噪声整形的语音去噪算法。该算法以最小感知均方误差为准则,在Wiener滤波的基础上,采用听觉感知加权函数修正Wiener滤波方程,实现对噪声谱整形,使噪声谱分布特性跟随语音谱而变;同时引入频率补偿因子克服非平稳噪声谱对语音影响的不均匀性;采用快速噪声估计算法实现对非平稳的估计。实验表明,该算法能更有效地抑制背景噪声,提高了去噪后的语音质量。
In this paper, a new speech denoising algorithm based on noise reshaping is proposed for non-stationary environments. The perceptual weighting function is adopted to modify the traditional Wiener filter equation via reshaping the noise spectrum based on the minimum perceptual mean square error criteria. The noise spectrum can be redistributed according to the real speech. The frequency compensation factor is introduced to compensate the ununiformity resulted f frequencies. The fast noise estimation rom alg that the algorithm can effectively reduc the orit e th in hm fluence of non-stationary noise on the speech spectrum at various is adopted to estimate the noise. The simulation result indicates e residual noise and improve the quality of the speech
出处
《通信技术》
2008年第12期253-255,258,共4页
Communications Technology
基金
国家自然科学基金资助项目(No:60572074)
关键词
噪声估计
非平稳信号
语音去噪
感知加权滤波器
噪声整形
Noise estimatien
Non-stationary signal
Speech denoising
Perceptual weighting filter
Noisereshaping
