Share Email Print

Proceedings Paper

Chinese dialect identification using prosodic classes and enhanced bigram model
Author(s): Linjia Sun
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

A method based on prosodic classes is proposed for Chinese dialect identification in this paper. The prosodic classes are obtained from a large number of prosodic words which are the basic unit of the prosodic structure, and simultaneously include acoustic, phonotactic and prosodic feature to classify dialects. In addition, the pauses between prosodic words also are considered and described as a special prosodic class. The different between the Chinese dialects is distinguished by the prosodic classes and their order in the whole sentences. The enhanced bigram model (EBM) based on HMM technique is proposed to obtain the sequential statistics of sequences of prosodic classes, which is shown to yields better identification performance and outperform the universal HMM model. We implement the new method to illustrate the capability of identification and evaluate it on the corpus from the Project for the Protection of Language Resources of China. The experimental results show that our method provides competitive performance with the existing methods.

Paper Details

Date Published: 27 November 2019
PDF: 6 pages
Proc. SPIE 11321, 2019 International Conference on Image and Video Processing, and Artificial Intelligence, 113212J (27 November 2019); doi: 10.1117/12.2539084
Show Author Affiliations
Linjia Sun, Beijing Language and Culture Univ. (China)
Ctr. for the Protection and Research of Language Resources of China (China)

Published in SPIE Proceedings Vol. 11321:
2019 International Conference on Image and Video Processing, and Artificial Intelligence
Ruidan Su, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?