Projet Parole
arrow Ressources Sphinx arrow Modèles acoustiques
Modèles acoustiques
Acoustic models (for CMU Sphinx 3.x decoder)

These models were estimated from manual transcriptions and audio records of radiophonic broadcast news (several radios, about 80 hours of training data).

Notice that these models don't work well with Sphinx 4. Why? Because they were estimated using SphinxTrain with the option -agc (automatic gain control) set to 'max'. At this time, Sphinx 4 does not propose this adjustment.
You can so use these models with Sphinx 4, but there will be not very reliable.


© 2018 Les outils du lium

Joomla! is Free Software released under the GNU/GPL License.