基于最小统计和人耳掩蔽特性的语音增强算法

Speech Enhancement Algorithm Based on Minimum Statistics and Human Auditory Masking Properties

提出了一种基于最小统计和人耳掩蔽特性的语音增强算法,通过最优平滑和最小约束递归平均从含噪语音中估计噪声的均值,推导出一种新的基于掩蔽特性的谱减系数计算公式。实验结果表明,该算法优于传统的掩蔽特性算法,含噪语音经过增强后,残留的音乐噪声更小。

A speech enhancement algorithm based on minimum statistics and human auditory masking properties is pro- posed. The mean vector of the noise spectrum is estimated from the noisy speech by optimal smoothing and minima controlled reeursive averaging. Then a novel estimation formula using masking properties for the spectral subtraction eoeffieients is giv- en. The experiment results show that the proposed algorithm is better than traditional masking algorithm and the enhanced speech has less musical noise.