Please use this identifier to cite or link to this item:
https://open.uns.ac.rs/handle/123456789/7609
Title: | A comparison of two prosody modelling approaches for Sesotho and Serbian | Authors: | Mohasi L. Sečujski, Milan Mak R. Niesler T. |
Issue Date: | 1-Jan-2014 | Journal: | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Abstract: | © Springer International Publishing Switzerland 2014. Accurate prediction of prosodic features is one of the critical tasks within a text-to-speech system, especially for under-resourced languages with complex lexical prosody. For synthesized speech to have a natural-sounding in-tonational contour, an adequate prosodic model should be employed. This study compares the Fujisaki model and the HMM-based prosodic modeling in the context of text-to-speech synthesis, for two quite distant languages with rich prosodic systems: Sesotho, a tonal language from the Bantu family, and Serbian, a South-Slavic language with pitch accent. The results of our experiments suggest that, for both languages, the Fujisaki model outperforms the HMM-based model in the modelling of the intonation contours of utterances of human speech. | URI: | https://open.uns.ac.rs/handle/123456789/7609 | ISBN: | 9783319115801 | ISSN: | 3029743 |
Appears in Collections: | FTN Publikacije/Publications |
Show full item record
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.