I’m working on a project for voice recognition on messages recorded on asterisk.
There is a huge difference on the quality of messages recorded when calling in from a GSM (mobile) phone and a softphone directly connected to asterisk.
The voice recognition program manages to understand more of the messages recorded from softphone clls than from GSM calls.
I know that GSM calls are uses less bandwidth than ulaw/alaw calls (13Kbps vs 64Kbps) so that the voice quality can never be the same.
But I would like to know what others have done to improve quality of recordings coming from GSM phones.
If you has worked on voice recognition I’d would be happy to hear about your experiences.