At first, MMSE algorithm is used for the speech enhacement module, and then, new robust features RAS-MFCC effectively suppressed the additive and convolutional noise.

• 该算法首先对带噪语音用最小均方误差估计法进行语音增强，后端对原始的带噪语音运用自相关法，以有效抑制加性和卷积噪声。