Re: Speaker Suggestions for Acoustic-Model Building (rif )


Subject: Re: Speaker Suggestions for Acoustic-Model Building
From:    rif  <rif(at)MIT.EDU>
Date:    Mon, 1 Nov 2004 08:40:15 -0500

> Sorry this isn't what you asked for, but I'm wondering if you > have considered the alternative of simply running your speech > files through room simulation or reverberation software, > with noise also added in software. > > This gives you lots of flexibility and you don't have to deal > with specific hardware issues. If and whan you want to > add those in, you can add them to your model separately. > You would only need to deal with these hardware issues > when you want a final demonstration of the working model > performance in a specific situation. > > > Just a thought... > > > Bob Masta This would be terrific if it were possible. I don't see how we'd do it --- we'd have to somehow model the exact noise characteristics of both our room and our microphone channel in a way that we could accurately simulate it. The room itself is reverberant, not especially regularly shaped, and there are several different known and somewhat non-stationary sources of nosie. Doing a simuation seems to be way more work than just recording a bunch of data through the channel, if it's doable at all. Cheers, rif


This message came from the mail archive
http://www.auditory.org/postings/2004/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University