计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (35): 101-105.

• 数据库、信号与信息处理 • 上一篇    下一篇

可训练语音合成在越南语合成中的应用

奎丽萍,杨  鉴,胡恩星,何  彬   

  1. 云南大学 信息学院,昆明 650091
  • 出版日期:2012-12-11 发布日期:2012-12-21

Appliance of trainable speech synthesis to Vietnamese

KUI Liping, YANG Jian, HU Enxing, HE Bin   

  1. School of Information, Yunnan University, Kunming 650091, China
  • Online:2012-12-11 Published:2012-12-21

摘要: 现代越南语是一种拼音化文字,采用“声、韵、调”标记其音节。采用基于HMM的可训练语音合成方法实现越南语语音合成。针对越南语的特点收集语料、录音、标注、确定音素列表、设计上下文属性集和问题集。在HTS平台下构建越南语的STRAIGHT语音合成器。对合成语音进行主观评测。初步评测结果表明,可懂度基本接近100%,自然度还处在“可以接受”和“比较自然”之间。

关键词: 越南语, 语音合成, STRAIGHT合成器, 主观评测

Abstract: The modern Vietnamese is a syllabic language. Each syllable can be marked with initial, final and tone. In this paper, Vietnamese speech synthesis is realized by applying a trainable HMM-based speech synthesis method. According to the characteristics of Vietnamese, it conducts such works as collecting corpus, recording, labeling, determining the phonemes list, designing context attributes and question set. The STRAIGHT synthesizer of Vietnamese speech is constructed under the HTS platform. It conducts a subjective test to synthetic speech signals. The results of preliminary evaluation show that intelligibility of the utterances is approximately 100%, and the quality of synthesis speech is from fair to good.

Key words: Vietnamese, speech synthesis, STRAIGHT synthesizer, subjective test