客語文句轉語音及語音辨認之研究

Abstract

本計畫擬進行客語語音處理之研究，包含兩個研究主題，一為建構一個以語料為基礎(Corpus-based)的客語文句轉語音 (Hakka Text-to-Speech, HTTS)系統，另一為建構一個實用的客語語音辨認系統，將包含四縣客語及海陸客語兩種個主要的客語腔調。研究內容包含：(1) 基本資料的收集及建購，如詞典、文字語料、語音語料等；(2) 客語構詞規則及文句分析；(3) 韻律模式之建立；(4) 客語Corpus-based TTS系統發展；(5) 語言模式之建立；(6) 客語語音辨認系統發展。

In this project, we should study two speech processing topics for Hakka language. One is the development of a high-quality corpus-based Hakka text-to-speech (HTTS) system. Another is the implementation of a Hakka speech recognizer. The study will cover the two main sub-dialects, 「Si-xian」 and 「Hai-lu」, of Kakka spoken in Taiwan. Several important issues will be studied. They include: (1) The collection and processing of lexicon, text databases and speech databases; (2) The realization of a high-performance text analyzer; (3) Prosody modeling; (4) The implementation of the HTTS system; (5) Language modeling; and (6) The realization of the Hakka speech recognizer.