Set acoustic model before loading speaker config in align

aalto-speech / AaltoASR

Aalto Automatic Speech Recognition tools

Other

85 stars 37 forks source link

Set acoustic model before loading speaker config in align #14

Closed fis closed 10 years ago

fis commented 10 years ago

Makes it possible to use model-based CMLLR also in conjunction with align.

fis commented 10 years ago

On closer look, they do seem to be equivalent; for consistency with e.g. phone_probs and mllr, perhaps the constructor approach would be better.

(I tried to avoid that initially just in case the model was expected to be initialized at that time, but that seems to be the way the other applications do it.)

psmit commented 10 years ago

Speaking of which, is there any file where it doesn't make sense to add the model directly to the SpeakerConfig?

Shall I change all constructors to include the model? @senarvi do you have an opinion?

psmit commented 10 years ago

I made a new pull request that puts the model in all appropriate speaker config constructors #15

psmit commented 10 years ago

Closed by #15