汉语连续语音切分技术研究  

Research on Segmentation of Chinese Continuous Speech

作　　者：王宇琛 张二华[1] WANG Yuchen;ZHANG Erhua(Nanjing University of Science and Technology,Nanjing 210094)

机构地区：[1]南京理工大学,南京210094

出　　处：《计算机与数字工程》2020年第8期1864-1869,共6页Computer & Digital Engineering

摘　　要：汉语连续语音切分的准确性,是影响汉语语音识别率的重要因素之一。连续语音切分是汉语语音识别的重要基础,论文介绍了汉语语音切分技术的基本原理,综合利用端点检测、基音周期轨迹、语谱图等特征,研究了一种汉语语音切分的算法。实验表明该算法能有效提高汉语连续语音切分的准确率。The accuracy of Chinese Continuous speech segmentation is one of the important factors affecting the quality of Chi⁃nese speech recognition.Continuous speech segmentation is an important foundation of Chinese speech recognition.This paper pres⁃ents an algorithm of automatic segmentation of Chinese speech by explaining and analyzing the basic principle of Chinese speech segmentation technology,based on the theory of endpoint detection,pitch track and spectrogram.The experimental results show that the algorithm can effectively improve the accuracy of Chinese speech segmentation.

关 键 词：语音切分 端点检测 语谱图 

分 类 号：H311[语言文字—英语]