2017 Fiscal Year Annual Research Report
Laryngeal mechanisms in speech production and perception: A cross-language study on F0 perturbation
| Project/Area Number |
17F17006
|
|---|
| Research Institution | Sophia University |
|---|
Principal Investigator |
荒井 隆行 上智大学, 理工学部, 教授 (80266072)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
GAO JIAYIN 上智大学, 理工学部, 外国人特別研究員
|
|---|
| Project Period (FY) |
2017-10-13 – 2020-03-31
|
| Keywords | Tokyo Japanese, / voicing, / pitch, / voice onset time, / electroglottograph, / laryngeal mechanism, / speech production, / speech perception |
|---|
| Outline of Annual Research Achievements |
[Voicing in Japanese] We focused our experiment on the acoustic realizations of voiceless and voiced stop consonants in Tokyo Japanese. We recorded two speakers for a pilot study and another eighteen speakers for the main experiment. Our main findings are: (1) a comprehensive report on the variations of the laryngeal timing relationship (or voice onset time) according to prosodic positions and pitch-accent contexts; (2) confirmation of recent findings (Takada et al. 2015) on the voiceless realizations of the voiced stops, suggesting a trend towards devoicing these stops; and (3) important and new findings: young Tokyo Japanese speakers are using remarkable pitch difference (on the following vowel) to distinguish voiceless and voiced stops. These results have been presented in the Spring Meeting of the Acoustical Society of Japan and will be presented in several international conferences in the summer, 2018. Besides voice onset time and pitch, detailed analyses of other acoustic parameters, such as duration and formant transitions, have been conducted.
[EGG recording] We tested the electroglottograph (EGG) device purchased recently and did recordings for the purpose of proposing new methods for speech synthesis. EGG signals were used as input for glottal pulses.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
We successfully investigated the phonetic realizations of consonantal voicing in different positions and contexts in Tokyo Japanese. We were thus able to show that Japanese had a different pattern from "true voicing", contrary to what has been proposed previously. Furthermore, the important pitch cue used by Japanese speakers in production confirms the tight relation between voicing and pitch, and gives insight on the understanding of tonogenesis (or tonal development) observed in many languages.
Analyses of other cues are being conducted and all these findings will be submitted to an international journal shortly. Meanwhile, we are preparing a perception test with Japanese listeners in order to investigate perceptual relevance of different acoustic cues of Japanese voiceless and voiced stops.
The EGG device has been successfully tested and future experiments will be conducted in order to observe the glottal opening state during the production of the voiceless and voiced consonants.
|
| Strategy for Future Research Activity |
[Perception of voicing in Japanese] Very shortly, we plan to investigate from the perceptual side how pitch and voicing are related in Tokyo Japanese, and furthermore, how perception and production are linked.
Since our production data in Tokyo Japanese showed very interesting and some unexpected results, we want to conduct a series of experiments by using other acoustic cues than pitch and voicing. By comparing results from Japanese listeners to those from, for example, Chinese or English listeners, we intend to determine which cues are universal and which cues are language-specific.
[Physiological study] After acoustic and perception tests, physiological studies are planned to investigate the laryngeal settings during the production of voiced and voiceless consonants, using EGG, fiberoptics, high-speed camera, etc.
[Improvement of Fujisaki model] The relevant cues used in production and perception will be proposed to improve the Fujisaki model, a tone and intonation model for pitch analyses and resynthesis.
|
