search

kaldi-asr / kaldi

kaldi-asr/kaldi is the official location of the Kaldi project.
http://kaldi-asr.org
Other
14.32k stars 5.33k forks source link

What is the procedure of training one far-distance ASR model for noise environment #4702

Open Sundy1219 opened 2 years ago

Sundy1219 commented 2 years ago

Hi, As we all know, recipes in kaldi is mostly for quiet environment. Now I want to train a ASR model with chain model for noise environment. my question is as follows: 1) Does i need to add noise segments to clean wav in GMM training stages? Namely, only clean wav can be used in GMM training stages, is right ? 2) the procedure of adding noise segments appeared in after GMM training and before chain model training，is right?. And, we need to use 'steps/align_fmmr_lats.sh' for realigning. is right ? Please tell me the whole procedure of training one far-distance ASR model with chain model for noise environment. Thank you

danpovey commented 2 years ago

It's probably not that critical when you add the noise. Search for scripts with 'multi_condition' in the name, these will be adding noise.

Sundy1219 commented 2 years ago

Thank you for reply. I see.

stale[bot] commented 2 years ago

This issue has been automatically marked as stale by a bot solely because it has not had recent activity. Please add any comment (simply 'ping' is enough) to prevent the issue from being closed for 60 more days if you believe it should be kept open.