F0 correlates of stress in Thai
An experiment was conducted to investigate changes in the fundamental frequency (F0) contours of Thai tones in connected speech as a function of stress. Thai has five tones: mid, low, falling, high, and rising. Stimuli consisted of 25 pairs of ambiguous target sentences with disambiguating context,...
Saved in:
Main Authors: | , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2024
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/179299 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | An experiment was conducted to investigate changes in the fundamental frequency (F0) contours of Thai tones in connected speech as a function of stress. Thai has five tones: mid, low, falling, high, and rising. Stimuli consisted of 25 pairs of ambiguous target sentences with disambiguating context, produced at a conversational speaking rate. One member of each pair contained a 2-syllable noun-verb sequence exhibiting a - -stress pattern, the other member a 2-syllable noun compound exhibiting a - stress pattern. Acoustic analysis revealed that Fo contours of stressed syllables more closely approximate Fo contours in citation forms than those of unstressed syllables. The degree of approximation is primarily determined by syllable structure and the interaction between adjacent tones. In contrast, FO contours of unstressed syllables undergo a more complex process. The average height of all five tones can be classified into three tonal registers: low, mid, and high. The low register comprises the low and the rising tones, the mid register the mid tone, and the high register the falling and the high tones. Based on shape, the falling and high tones are distinguished within the high register, the low and rising tones within the low register. Therefore, a five-way contrast among all five tones appears to be maintained in both stressed and unstressed syllables. In addition, two statistical parameters, average Fo and coefficient of variation, are proposed for a machine model to automatically detect stressed and unstressed syllables. |
---|