Journal of the Chinese Institute of Engineers, Volume 22, No. 5, September 1999.211 Transactions of the Chinese Institute of Engineers, Series A. Special Issue on 211 Chinese Speech and Language Processing

摘要

This paper presents a semi-automatic phonetic labeling method for processing in211u001ethe MAT (Mandarin across Taiwan) speech database. MAT speech data are collected 211u001ethrough the telephone networks. Each utterance has been transcribed into Chinese 211u001echaracters and Pinyin symbols. The proposed phonetic labeling method will mark 211u001ethe syllable and sub-syllable boundaries in an utterance. Phonetic symbols are 211u001eassigned to each segmented syllable. The segmentation process is accomplished by 211u001eusing hidden Markov modeling (HMM) and Viterbi decoding. The accuracy of syllable 211u001esegmentation is detected by measuring the syllable length and the distance of a 211u001esyllable from its state models. The experiment results show that the proposed 211u001elabeling method can achieve segmentation accuracy around 90% for an allowed 211u001etolerance of 16 ms.