利用线性预测残差的语音去噪方法

Denosing of speech signal using linear prediction residual

语音信号中经常混有加性噪声,噪声的存在会使语音处理系统的性能急剧下降,因此语音去噪是语音信号处理中的关键技术之一.提出一种利用线性预测残差去除语音中的加性白噪声的方法,实验结果表明,该方法可以有效地去除语音信号中的加性白噪声,并且保持语音信号较好的清晰度和可懂度.

A speech signal is frequently mixed with the additive noise. Due to the existences of the noise, the performances of the speech processing system will decline remarkably, so densising is one of the key techniques in speech signal processing. In this paper method is proposed for removing the additive white noise in the speech by using linear predication residual. Experiment results show that this method can remove the additive white noise in the speech signal effectively, while remain good the articulation and intelligibility.