提出了一种基于WDFT的PLP特征提取算法。该算法采用WDFT来提高语音信号低频部分的频谱分辨率,与传统的提取方法相比更符合人类听觉系统的特性。对0～9十个英文数字和E-SET集的识别实验结果表明,本文提出的特征提取算法比传统PLP算法具有更好的识别效果。 A novel feature extraction algorithm for PLP based on WDFT is proposed. This algorithm uses WDFT to improve the spectral resolution of the low frequency part of the speech signal, which is more in line with the characteristics of the human auditory system than the traditional extraction method. Experimental results on 0 ~ 9 ten English digits and E-SET set show that the proposed feature extraction algorithm has better recognition performance than the traditional PLP algorithm.