2022 Fiscal Year Annual Research Report
Psychoacoustic roughness as a measure of glottalization in consonants
| Project/Area Number |
19K13162
|
|---|
| Research Institution | The University of Aizu |
|---|
Principal Investigator |
Perkins Jeremy 会津大学, コンピュータ理工学部, 上級准教授 (30725635)
|
|---|
| Project Period (FY) |
2019-04-01 – 2023-03-31
|
| Keywords | roughness / f0 / spectral tilt / Korean / Thai / laryngeal constriction |
|---|
| Outline of Annual Research Achievements |
This project was concluded successfully in 2022, with Korean data collected and analyzed in summer 2022. Two papers were then written from these results during the winter of 2022-23. Both papers were accepted for publication and presentation at ICPhS 2023 in Prague, Czechia.
The first paper was an acoustic analysis of the plain and tense fricatives, showing that release duration and spectral tilt differed; also, psychoacoustic roughness was analyzed, showing that it performed equally well as spectral tilt in identifying Korean tense obstruents.
The second paper was written jointly with Prof. Yu Yan, a Computer Scientist whose knowledge on machine learning allowed us to use the random forest machine learning technique to test which acoustic cues were necessary and sufficient to identify the three Korean obstruent types (tense, lax and aspirated). This paper offers additional insights into which acoustic measures were most important in the learning algorithm to successfully learn the differences between the three Korean obstruents. These importance values can offer a possible insight into which of these acoustic cues may be most useful in phonological perception.
|
Research Products
(3 results)
