Please use this identifier to cite or link to this item:
http://hdl.handle.net/10397/114601
Title: | Key acoustic cues for the realization of metrical prominence in tone languages : a cross-dialect study | Authors: | Hu, Y Feng, H |
Issue Date: | 2024 | Source: | Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, 2024, p. 3130-3134 | Abstract: | Few studies have quantitatively discussed the acoustic realization of metrical prominence across tone languages. This study addresses this issue in word-level prosodic units, aiming to identify effective acoustic cues for cross-language analysis based on the realization of metrical prominence. Findings include: (1) Within-dialect analyses reveal robust acoustic cues for realizing metrically strong units: duration, pitch height, and pitch slope for both left- and right-dominant prominence, and intensity, pitch range and cumulative pitch dynamics, an often-overlooked cue, for right-dominant one; (2) Cross-dialect analyses by Boruta algorithm identify that all these six acoustic cues are significant for the classification of Chinese dialects according to their realization of metrical prominence. Linear Discriminant Analysis emphasizes duration, intensity, and cumulative pitch dynamics as key discriminators across dialects. | Keywords: | Acoustic realization Chinese dialects Machine learning Metrical prominence |
Publisher: | International Speech Communication Association | DOI: | 10.21437/Interspeech.2024-666 | Description: | Interspeech 2024, 1-5 September 2024, Kos, Greece | Rights: | The following publication Hu, Y., Feng, H. (2024) Key Acoustic Cues for the Realization of Metrical Prominence in Tone Languages: A Cross-Dialect Study. Proc. Interspeech 2024, 3130-3134 is available at https://doi.org/10.21437/Interspeech.2024-666. |
Appears in Collections: | Conference Paper |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
hu24_interspeech.pdf | 1.05 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.