论文部分内容阅读
对语音信号进行处理一个很重要的问题就是噪声的滤除,噪声降低了语音的信噪比和可懂性。多年来,人们针对加性宽带噪声提出了各种语音增强算法,其中谱减法因其具有简单和易于实现的优点而被广泛的采用。但是,一般的谱减法都会带来较为严重的“音乐”噪声,为减弱这种噪声,本文提出了一种改进的语音增强算法——多带谱减法(multi-bandspectrumsubtraction)。实验结果表明,多带谱减法明显的削弱了一般谱减法所带来的音乐噪声,对带噪语音质量的增强效果显著。
A very important issue when dealing with speech signals is noise filtering, which reduces the signal-noise ratio and intelligibility of speech. Over the years, various speech enhancement algorithms have been proposed for additive wideband noise. Spectral subtraction has been widely adopted because of its simplicity and ease of implementation. However, the general spectral subtraction will bring more serious “music” noise. In order to reduce this kind of noise, an improved speech enhancement algorithm named multi-bandspectrumtraction is proposed. The experimental results show that the multispectral subtraction significantly weakens the music noise brought by the general spectral subtraction, and significantly enhances the quality of the noisy speech.