14th International Congress of Phonetic Sciences (ICPhS-14)

San Francisco, CA, USA
August 1-7, 1999

A Phonetic Labeling Method based on Hidden Markov Models and Viterbi Decoding Procedure

Hsiao-Chuan Wang, Rong-Liang Chiou, Shiang-Kai Chuang, Yi-Fen Huang

National Tsing Hua University, Hsinchu, Taiwan

This paper presents a phonetic labeling method for the processing of Mandarin speech database produced in Taiwan. These speech data are collected through the telephone networks. Each utterance has been transcribed into Chinese characters and Pinyin sequence. The proposed phonetic labeling method will mark the syllable boundaries in an utterance and assign the phonetic symbols to every syllable. Each syllable is partitioned into two sub-syllables, i.e. the initial and the final. These sub-syllables are represented in hidden Markov models. The labeling process is accomplished by using Viterbi decoding procedure. The labeling correctness is also automatically examined.

Full Paper

Bibliographic reference.  Wang, Hsiao-Chuan / Chiou, Rong-Liang / Chuang, Shiang-Kai / Huang, Yi-Fen (1999): "A phonetic labeling method based on hidden Markov models and Viterbi decoding procedure", In ICPhS-14, 1701-1704.