Task #4251
closedTask #3678: RA3c - Continuity of prosodic patterns
Task #4237: Continuity of F0 pattern
Task #4250: F0 join cost
Propose F0 weighting for F0 continuity in phones
100%
Description
Propose F0 weighting F0 continuity in phones.
The idea:
F0 continuity is important only when vowels (or also other sonorants?) are concatenated. For other phones, F0 continuity is not so important or even can be ignored.
Propose a phone-level weighting scheme (the bigger the weight, the more important the F0 continuity is).
Files
Updated by Skarnitzl Radek over 7 years ago
- Due date changed from 31.10.2017 to 01.12.2017
- % Done changed from 0 to 20
We are preparing stimuli for a controlled experiment. VCV pseudowords have been monotonized in terms of their F0 and, subsequently, F0 is being step-wise modified in the middle of the target consonant to emulate a sudden F0 shift in the point of concatenation.
Updated by Skarnitzl Radek over 6 years ago
The results of the perceptual experiment clearly show that the current practice of including F0 in all voiced sounds as a criterion in calculating the concatenation cost is not necessary. The study is prepared for submission into a journal (and should still be published this year); the results are summarized as follows:
1) The direction of F0 change needs to be taken into account; in the attached figure, the discontinuity in the two F0 courses is objectively the same, but the first one will not be audible, while the second will be disruptive.
2) F0 can be ignored within obstruent consonants (i.e., plosives, fricatives, affricates); only calculate F0 when concatenating diphones pertaining to sonorants (nasals, approximants).
3) F0 discontinuities (in sonorants) of 1 semitone or less may be ignored; in the test, we had discontinuities of 1 or 5 ST, the former were never perceived as disruptive. Future experiments may focus on where between 1 and 5 ST the boundary lies.
Updated by Matoušek Jindřich about 6 years ago
- Assignee changed from Skarnitzl Radek to Tihelka Dan