提出了一种基于线性逻辑回归的方法,对利用基频和MFCC特征获得的分数进行融合来进行说话人的性别识别,其中包括了基于基频特征的单高斯模型和基于MFCC特征的混合高斯模型.采用语音库包括男性语音文件150个,女性语音文件190个.实验结果中识别率可高达97.65%,比传统单用基频或是MFCC特征的识别率都要高,具有更好的判别性能. A method based on linear logistic regression was proposed to fuse the speaker’s gender by fusion of the scores obtained from the fundamental frequency and the MFCC features, including single-Gaussian model based on fundamental frequency features and Gaussian mixture model based on MFCC The speech database includes 150 male voice files and 190 female voice files.The recognition rate in the experimental results can reach as high as 97.65%, which is higher than the recognition rate of the traditional single-frequency or MFCC features and has better discrimination performance .