Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!agate!ames!aio!ROIL1!john
From: john@aio.jsc.nasa.gov  (John Salmon)
Subject: Re: Logistics of demonstrating Speech applications?
Message-ID: <940210105620@ROIL1>
Nntp-Software: PC/TCP NNTP
Lines: 32        
Sender: usenet@aio.jsc.nasa.gov (USENET News Client)
Reply-To: salmon@smsd.jsc.nasa.gov
Organization: nasa-jsc
References: <1994Feb8.193858.1059@sol.cs.wmich.edu> <CKyvHH.40F@lut.ac.uk>
Date: Thu, 10 Feb 1994 10:56:20

In article <CKyvHH.40F@lut.ac.uk> S.Datta@lut.ac.uk writes:

    In article <1994Feb8.193858.1059@sol.cs.wmich.edu> parens@sol.cs.wmich.edu (Put Your Name Here) writes:
    >Well congratulate me, I just gave my first demo of a speech navigated
    >application.  The application showed well, but there's a problem,
    >because how do you *talk* about the system without the system responding
    >to your sales pitch.  
    >
    >In other words, what is the best microphone gear to use for this type of
    >thing.  I assume some sort of uni-directional microphone head piece that
    >is switch activated, so you can talk to the people, then talk to the 
    >system.  
    >
    >Any wisdom is this area?
    >.
    >
    
    I remember implementing a system some twenty years back for interrogating 
    a Relational Database by Voice. The input was in the form of isolated 
    words, but using the telephone handset as the transducer. I encountered 
    a similar problem to the one you are referring. The solution I opted for 
    was to introduce two control words STOP and GO. Whenever the system heard 
    the word STOP it suspended all activities except to listen continually 
    for the word GO for reactivation. I made sure that these control words 
    needed to be uttered fairly distinctly for the system to perform 
    satisfactorily.
    
    
    Sekhar Datta.
    
Use a microphone with an on-off switch. You may have to make the system
ignore noise generated by the switch.

