Hi, I have my application running in spanish and it performs very well when used with very hi-quality microphones and a console, however when used with the laptop microphone the accuracy decreases(its not bad though).
Of course this is a logic thing to happen but it would be great to make my application a bit more independent in regard to the incoming audio.
Not asking for a miracle of course, but i wonder if there are certain "tunes up " that could be made in order to increase accuracy.
Thanks,
Santiago
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
> Of course this is a logic thing to happen but it would be great to make my application a bit more independent in regard to the incoming audio.
Train model on noise data or adapt it either offline or online. Create a noise cancellation preprocessor. It's hard to expect high accuracy from the current code.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi, I have my application running in spanish and it performs very well when used with very hi-quality microphones and a console, however when used with the laptop microphone the accuracy decreases(its not bad though).
Of course this is a logic thing to happen but it would be great to make my application a bit more independent in regard to the incoming audio.
Not asking for a miracle of course, but i wonder if there are certain "tunes up " that could be made in order to increase accuracy.
Thanks,
Santiago
> Train model on noise data
Very interesting, where can i find info about this, please try to be precise cause i tend to get lost looking uo for info in the API...
>or adapt it either offline or online
What is this??
>Create a noise cancellation preprocessor
Is this a module of sphinx project or you meant to say i should design an external audio filter?
Thank you very much, i apologize for asking so many quiestions,
Santiago
> Very interesting, where can i find info about this, please try to be precise cause i tend to get lost looking uo for info in the API...
http://www.speech.cs.cmu.edu/sphinx/tutorial.html
> What is this??
http://www.speech.cs.cmu.edu/cmusphinx/moinmoin/AcousticModelAdaptation
> Is this a module of sphinx project or you meant to say i should design an external audio filter?
It doesn't matter. This module can be either external or internal. It's purpose is to clean noise. Something like this
http://portal.acm.org/citation.cfm?id=1170742.1170828
would be nice to implement
> Of course this is a logic thing to happen but it would be great to make my application a bit more independent in regard to the incoming audio.
Train model on noise data or adapt it either offline or online. Create a noise cancellation preprocessor. It's hard to expect high accuracy from the current code.