Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!nntp.sei.cmu.edu!cis.ohio-state.edu!math.ohio-state.edu!jussieu.fr!oleane!tank.news.pipex.net!pipex!uknet!info!myhost.subdomain.domain!iijohn
From: iijohn@myhost.subdomain.domain (John Openshaw)
Subject: Re: Speech Recognition in Cockpit?
X-Nntp-Posting-Host: iifeak.swan.ac.uk
Message-ID: <DCoFqG.F82@info.swan.ac.uk>
Sender: news@info.swan.ac.uk
Organization: i2it
X-Newsreader: TIN [version 1.2 PL2]
References: <3u2vms$ddn@eccdb1.pms.ford.com> <3v719h$28d@newsbf02.news.aol.com>
Date: Wed, 2 Aug 1995 09:30:15 GMT
Lines: 28

M Pender (mpender@aol.com) wrote:
: If you think an aircraft cockpit is much quieter than a car, 
: you've never been in an aircraft cockpit.

Thanks for that useful comment. Just to re-iterate a few points that I
think were made here recently, concerning recognition in the two 
environments. In an aircraft cockpit you have a highly trained speaker
using a highly trained speaker dependent system. He doesnt mind having
an intrusive face-mask to pick up his speech and shield the mic from some
of the ambient noise. Aircraft noise compared to car noise is relatively
stable - its one thing getting robust recognition with a fixed SNR Gaussian
noise model, but when the noise spectrum and level change in time, it
starts to get a little more involved.

So an aircraft cockpit may be louder, but in terms of the actual recognition
environment things are trickier in a car.


John Openshaw	e-mail: j.p.openshaw@swansea.ac.uk









