藏语音素转写的算法研究
Research on Tibetan Grapheme to Phoneme Letter to Transformation Phoneme Sequence Algorithm for Speech Synthesis
拉巴顿珠 1珠杰 1顿珠次仁1
作者信息
- 1. 西藏大学信息科学技术学院,西藏 拉萨 850000
- 折叠
摘要
藏语语音合成技术中,字-音转换是一项必不可少的关键性研究任务,即将藏语音节转写成对应的音素序列.从藏语语音合成技术的任务出发,以藏语传统文法为依据,分析其现代藏语音节的拼写结构和连续文本的复杂特征,提出一种基于拼写结构的藏语音素转写算法.经测试表明,只要是拼写规范的音节,则音素转写准确率高达 99%以上.另外,所提出的音素转写方案不仅可以有效地处理语音合成技术中的字-音转换任务,而且对藏文排序、藏语文教学以及藏文信息处理的其它研发任务具有一定的参考意义,对于规范藏语、简化藏文编码等起到显而易见的作用.
Abstract
In Tibetan speech synthesis technology,word-to-speech conversion is an essential and key research task,that is,Tibetan syllables are converted into corresponding phoneme sequences.Starting from the task of Tibetan speech synthesis technology and based on traditional Tibetan grammar,this paper analyzes the spelling structure of modern Tibetan syllables and the complex characteristics of continuous text,and proposes a Tibetan phoneme tran-scription algorithm based on spelling structure.The test shows that as long as it is a syllable with standard spelling,the accuracy of phoneme transcription is more than 99%.In addition,the phoneme conversion scheme proposed in this paper can not only effectively deal with the task of word-to-sound conversion in speech synthesis technology,but also has a certain reference significance for Tibetan sorting,Tibetan language teaching and other research and development tasks of Tibetan information processing,and plays an obvious role in standardizing Tibetan and simplifying Tibetan coding.
关键词
语音合成/拉丁字母/音素/藏语Key words
Speech synthesis/Latin alphabet/Phonemes/Tibetan引用本文复制引用
基金项目
教育部人文社会科学研究项目(21YJCZH059)
教育部人文社会科学研究项目(20YJA740035)
西藏大学培育计划项目(ZDCZJH21-10)
出版年
2024
NETL
NSTL
万方数据