Browse Prior Art Database

Method for Switching a Speech Recognizer from Asynchronous to Synchronous Recognition

IP.com Disclosure Number: IPCOM000104099D
Original Publication Date: 1993-Mar-01
Included in the Prior Art Database: 2005-Mar-18
Document File: 2 page(s) / 82K

Publishing Venue

IBM

Related People

Daggett, G: AUTHOR [+3]

Abstract

Disclosed is a method of switching a speech recognition system between asynchronous and synchronous modes.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Method for Switching a Speech Recognizer from Asynchronous to Synchronous Recognition

      Disclosed is a method of switching a speech recognition system
between asynchronous and synchronous modes.

      There  are  two  basic  modes of operation for a speech
recognizer [*]:  synchronous and asynchronous.  In asynchronous or
"dictation" mode, the recognizer continuously  processes speech until
the microphone is turned off.  Words are displayed at the workstation
as soon as they are decoded.  This mode is usually used when
dictating free-flowing text to the recognizer while the vocabulary
and  language  model remain unchanged.  In synchronous or "command"
mode, a single word or phrase is decoded  only  upon  receipt  of  a
request from the controlling application.  This mode is used when
issuing  commands to an application which may cause it to step
through a series of  menus, possibly  changing the active  vocabulary
after receiving a word and before issuing a request to decode the
next word.

      Switching between these modes can be accomplished by some
keyboard operation, but it is more natural if it is accomplished by
voice command.  Switching out of command mode is simple since the
workstation is in full control and must  merely  issue  a  request
to the recognizer to change mode.  Switching out of dictation mode is
less simple, since the verbal command must be distinguished from the
dictated text, and the workstation must be notified that the
recognizer is now in synchronous mode.

      Another case is when the vocabulary must be changed to one of
the letters and digits, so that a word can be spelled out.  The word
may be a rare one, outside the active vocabu- lary,  or  it  could
be  a sequence of characters such as a license plate identifier.
This  can  be  handled  entirely within  the  asynchronous  mode  by
defining a word known to both the workstation and the recognizer
that  performs  the vocabulary switching.  But this is an inflexible
solution as all  vocabularies  and  words to switch between them must
be predefined to both the recognizer and the controlling  work-
station application.

      The  method  presented  here  allows  the recognizer to switch
into asynchronous mode whenever it decodes one  of  a set  of  verbal
commands, referred to as "stop words".  The word is sent to the
workstation along with a flag  to  indi-...