摘要： 证实普通话可以分解为辅音音素和单元音音素通过过度音的连接，提出一种单字音特征提取方法。该方法在传统的帧特征提取基础上，对相关帧进行二次处理，得到单字语音中的多个代表帧，将代表帧进行拼接作为单字的特征矢量。这种特征提取方法能更好地表现说话人单字发音中相邻语音帧之间的连续性。仿真实验表明该方法在说话人识别系统的应用中达到较高的识别率，使识别时间进一步缩短。

关键词: 说话人识别, 特征提取, 单字音特征

Abstract: In this paper, a novel method of extracting speaker character is proposed, which is based on the single word. Compared with conventional methods, this method aims at interrelated frame, effectively finding multi-representative frames in single word utterance, and puts together the character vectors of these frames to constitute the character vector of this single word, which is adopted to calculate the codebook and model matching in training and recognition process. Simulation experiment indicates that this method can acquire higher recognition rate in Speaker Recognition(SR) system.

Key words: Speaker Recognition(SR), character extracting, single word character

中图分类号: 

• TN912.34