摘要
根据汉语语音信号的特点．该文提出一种简化的自相关基音提取的方法，算法具有简便、快速、准确和易于实现等特点。此方法不仅可用于孤立字．更适用于提取二字词的声词。实验表明，所得到的１５种二字词声调的基本模式和轻声变调等都与已有的研究报导相符。含有清声母音节时，二字词声调能提供音节分割和声韵分离的准确信号．基音提取成功率高于９０％。
This paper presents a simplified autocorrelation method for pitch for the Chinese tone hased on the analysis of tone pattern for mandrin double-syllabic word. The algorithm is characterized by its simpleness, high speed.accuracy and easy realization.The 15 fundamental tone pattems and neutral tone for double-syllablic word obtained in our experiment is correspond-ing to the other reported results.If the double-syllabic word contains the voiceless initial, then the tone pattern can carry the information for syllable segmentation and the separation of the initial and final.The accuracy of pitch detection is above 90%. The main error exists in some of the 4th tone for double-syllabic word and is caused by the intense response of vocal tract.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
1996年第3期31-36,共6页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金
关键词
语音分析
基音提取
自相关计算
汉语
算法
speech analysis, pitch detetion,autocorrelation calculation
