Activity
From 23.08.2017 to 21.09.2017
21.09.2017
- 10:16 Task #4258 (Closed): Test the consistency of P0 prosodemes in our speech corpora
- *The idea:*
P0 prosodemes should have "flat" (or "neutral") intonation when compared to other prosodemes. Using appr... - 10:06 Task #4257 (Closed): Introduce a special prosodeme for utterance-leading prosodic word
- Introduce a special prosodeme for utterance-leading prosodic word (similarly as it is done for utterance-trailing pro...
- 09:59 Task #4256 (Closed): Revise prosodic word parsing of 'wh' questions
- The current rule-based of 'wh' questions is a [[Wh_Questions|compromise]]. Please try to revise it and propose a new ...
- 09:54 Task #4255 (Closed): Revise prosodic word parsing
- The current rule-based [[Prosodic_words|prosodic word parsing]] of an utterance is not perfect. Please try to revise ...
- 09:16 Task #4254 (Closed): Prosodic word based positional features
- _Parent task for experiments with positional features related to prosodic words_
- 09:08 Task #4253 (Closed): Set up unit selection to ensure that 2nd syllable has higher F0 than 1st syllable in non-final prosodic words
- Based on the analysis in #4252, propose an experiment to ensure that the 2nd syllable has higher F0 than the 1st syll...
- 09:00 Task #4252 (Closed): Analyze F0 differences in first two syllables of non-final prosodic words
- F0 in the 2nd syllable should be higher than F0 in the 1st syllable.
Check it out on several recorded corpora. - 08:51 Task #4251 (Closed): Propose F0 weighting for F0 continuity in phones
- Propose [[F0_concatenation_cost|F0 weighting]] F0 continuity in phones.
*The idea:*
F0 continuity is important on... - 08:32 Task #4250 (New): F0 join cost
- _Parent task for experiments with F0 join cost computation_
- 08:25 Task #4249 (Closed): Spectral tilt for join cost
- Consider spectral tilt in join cost computation.
Compare MFCC- and formant-based join cost with the spectral tilt ... - 08:13 Task #4248 (New): Data-based context penalization matrix
- Propose data-based context penalization matrix.
*The idea:*
Use a similarity matrix in the range <0; 1> instead o... - 08:05 Task #4247 (Postponed): Expert-based context penalization matrix
- Revise and propose expert-based [[Context_penalty|context penalization matrix]].
18.09.2017
- 10:36 Task #4205: DNN-based outlier duration detection and penalization
- Experiments with a new feature in target cost - difference of real duration of current unit and the predicted value (...
- 10:36 Task #4206: HMM-based outlier duration detection and penalization
- Experiments with a new feature in target cost - difference of real duration of current unit and the predicted value (...
- 08:43 Task #3944 (Closed): Commit changes based on user(s) logged in/out
- OK. I have no problem with closing it when it is too difficult. I just wanted to avoid a partial commit, when one use...
- 07:41 Task #4212 (Resolved): Correlation of EMA-based articulation data with MFCCs
- Correlation between EMA data and MFCCs was examined for *one*, *ten* and *a hundred* different sentences. Resulting c...
- 07:22 Task #4095 (Resolved): Synthetize sentences with new features
- Different combinations of articulation data were used as features, to synthetize sentences with voice of Radka Maskov...
17.09.2017
- 20:11 Task #4139 (Assigned): Fix segmentation of pauses
- 20:10 Task #3944: Commit changes based on user(s) logged in/out
- According to Jakub, this is not an easy job. I move this issue back to Dan - please revise it and possibly close the ...
- 20:07 Task #3884 (Closed): Fix the order of alternative candidates
- 20:06 Task #3883 (Closed): Nastavení - page unavailable
- 20:05 Task #3923 (Closed): Add frequency grid to spectrogram
16.09.2017
- 22:31 Task #4139: Fix segmentation of pauses
- It is technically finished. Some testing and tuning could be still performed (but maybe it is not necessary).
- 22:24 Task #4211 (Closed): Praat script to compute spectral slopes
15.09.2017
- 14:57 Task #3944 (Resolved): Commit changes based on user(s) logged in/out
- I don't think that this is necessary. It is hard to detect when user closes up the browser. The current state is bett...
- 14:55 Task #3923: Add frequency grid to spectrogram
- I think this is sufficient. I am closing this issue.
- 14:53 Task #3923 (Resolved): Add frequency grid to spectrogram
- 14:54 Task #3884 (Resolved): Fix the order of alternative candidates
- 14:54 Task #3883 (Resolved): Nastavení - page unavailable
- passwords can be changed now
13.09.2017
- 19:00 Task #4244 (Closed): Classifier spot checking
- # Check various classifiers:
** linear
** non-linear
** ensemble
** MLP
# Select the best classifier (coarse gri... - 18:50 Task #4242 (Closed): Context size for current GCI detection
- Try various numbers of peaks around the current peak candidate
- 18:41 Task #4241 (Closed): New acoustic features
- * harmonic/noise
* voiced/unvoiced (V/U)
* F0 contour
- 18:39 Task #4240 (Closed): Implement GCI evaluation metrics
- Done by Daniel Tihelka.
- 18:38 Task #4240 (Closed): Implement GCI evaluation metrics
- * Metrics:
** IDR
** missing rate
** false alarm rate
** accuracy (A25)
** IDA
** combined evaluation
* Python - 17:53 Task #4239 (Closed): Baseline GCI detector
- Published: "MATOUŠEK, J., TIHELKA, D. Classification-Based Detection of Glottal Closure Instants from Speech Signals"...
- 17:48 Task #4239 (Closed): Baseline GCI detector
- * based on standard classification techniques
* features inspired by (Barnard1991)
- 17:53 Task #4238 (Assigned): Classification-based GCI detection
- 17:51 Task #4238 (Closed): Classification-based GCI detection
- 17:43 Task #4238 (Closed): Classification-based GCI detection
- _Parent task for classification-based glottal closure instant (GCI) detection_
- 17:40 Task #4237 (New): Continuity of F0 pattern
- _Parent task for the continuity of the F0 pattern within the synthesized utterance_
- 13:00 Task #3690 (Closed): Annotation error detection
Also available in: Atom