【摘要】 <正>1.引言 当前有很多语音识别系统和产品,但绝大部分是工作衣安静环境下的,一旦在噪环境下使用,语音信号中混人严重的背景噪声,信噪比就大为下降,影响了参数的稳定性,而通常采用的语音特征在更多还原

【Abstract】 In this paper, we review the noise suppression methods in the speech recognition, and divide the methods into four major areas: removing noise estimation, environment normalization technique, updating model technique, and building noise model. We can not review all possible techniques of noise suppression in speech recognition, because there are far too many methods in use today, Instead, we only review those techniques most commonly used.更多还原