基于归一化自相关的语音基频特征提取  

Fundamental Frequency Feature Extraction of Speech Based on the Normalized Cross Correlation Function

作　　者：曹梦霞[1] 郑永果[1] 郑尚新 

机构地区：[1]山东科技大学,山东青岛266590

出　　处：《信息技术与信息化》2014年第2期49-51,共3页Information Technology and Informatization

摘　　要：语音的基频(也称音高、基音周期或F0)及其变化规律是语音信号的一个重要特征,在语音情绪识别、声纹识别中有重要的应用。而语音基频的提取一直是语音信号处理中的难点,这也是语音基频特征未能广泛应用于语音识别等应用的重要原因,因此准确高效的提取音高在语音信号处理中能够有重要的意义。本文基于归一化自相关函数,结合倒频谱方法,提取了一种改进的基于归一化自相关的语音基频提取算法,实验证明该方法在基频提取中取得了较好的结果。The fundamental frequency （also called Pitch or F0） and its change rule are important features of speech signals for speech emotion recognition and voice print recognition and so on. But fundamental frequency extraction has always been the difficulty in the speech signal processing field, and the reason F0 not widely used in various applied in speech recognition is that so accurate and efifcient to extract the pitch can have important signiifcance in the speech signal processing. Based on the normalized cross correlation function, combining the cepstrum method, a new fundamental frequency extraction algorithm is proposed, and the experiment proved that the method in the fundamental frequency extraction has achieved good results.

关 键 词：音高 F0 基频 归一化自相关 语音 信号处理 

分 类 号：TN912.3[电子电信—通信与信息系统]