在本文中，我们提出一种噪声环境下具有鲁棒性的语音特征提取的新方法。首先介绍了在高噪声环境下语音识别所遇到的特殊问题，随后对现有的特征提取的方法做了比较，提出了基于线性预测的ＭＦＣＣ特征提取方法，通过实验比较证明该方法具有较好的鲁棒性 In this paper, we propose a new method of robust speech feature extraction in noisy environments. Firstly, the special problems encountered in speech recognition in high-noise environments are introduced. Then, the existing methods of feature extraction are compared, and the MFCC feature extraction method based on linear prediction is proposed. The experimental results show that the proposed method is better Robustness