论文部分内容阅读
提出两种基于非对称代价函数的耳语音增强算法,将语音增强过程中的放大失真和压缩失真区分对待。Modified ItakuraSaito(MIS)算法对放大失真给予更多的惩罚,而Kullback-Leibler(KL)算法则对压缩失真给予更多的惩罚。实验结果表明,在低于—6 dB的低信噪比情况中,经MIS算法增强后的耳语音的可懂度相比传统算法有显著提高;而KL算法则获得了同最小均方误差语音增强算法近似的可懂度提高效果,证实了耳语音中的放大失真和压缩失真对于耳语音可懂度的影响并不相同,低信噪比时较大的压缩失真有助于提高耳语音可懂度,而高信噪比时的压缩失真对耳语音可懂度影响较小。
Two kinds of ear speech enhancement algorithms based on asymmetric cost function are proposed, which treat the amplification distortion and compression distortion in speech enhancement process. The Modified ItakuraSaito (MIS) algorithm penalizes for increased distortion, while the Kullback-Leibler (KL) algorithm penalizes compression distortion more. The experimental results show that the intelligibility of ear speech enhanced by MIS algorithm is significantly higher than that of the traditional algorithm at low signal-to-noise ratio of less than -6 dB, while the KL algorithm obtains the same mean-square error speech Enhancement of the algorithm to enhance the intelligibility of the approximation algorithm proves that the amplification distortion and compression distortion in the ear speech have different effects on the speech intelligibility of the ear and the larger compression distortion at the low signal-to- Comprehension, while the high signal-to-noise ratio of the compression distortion of the audible tone less affected.