Activity
From 29.10.2017 to 27.11.2017
27.11.2017
- 13:56 Task #4249: Spectral tilt for join cost
- I have just started experiment with:
* spectral tilt (static) + MFCC (classic)
* spectral tilt (static) + formants ...
15.11.2017
- 18:32 Task #4241 (Closed): New acoustic features
- New acoustic, spectral and MFCC features tried. In general, they improve prediction accuracy. The best results were o...
10.11.2017
- 09:09 Task #3965 (Closed): EMA - Radka Mašková
- 09:09 Task #4095 (Closed): Synthetize sentences with new features
- 09:09 Task #4212 (Closed): Correlation of EMA-based articulation data with MFCCs
- 08:49 Task #4242 (Closed): Context size for current GCI detection
- Tested for extremely randomized trees (ERT) classifier => 3 preceding and 3 suceeding peaks seem to be the best choice
07.11.2017
- 08:37 Task #4252: Analyze F0 differences in first two syllables of non-final prosodic words
- Updated files (with beg times of vowels) are available on page [[F0_difference]], the structure of the files is now a...
06.11.2017
- 19:32 Task #4251: Propose F0 weighting for F0 continuity in phones
- We are preparing stimuli for a controlled experiment. VCV pseudowords have been monotonized in terms of their F0 and,...
31.10.2017
- 09:21 Task #4271 (Postponed): Detection on 8kHz-based features
- Downsampling to 8kHz does not seem to be so much important at this moment => postponing this issue...
Also available in: Atom