Browse Prior Art Database

Method to Pace User Input for Speech Recognition Systems

IP.com Disclosure Number: IPCOM000113979D
Original Publication Date: 1994-Oct-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 4 page(s) / 123K

Publishing Venue

IBM

Related People

Baez, D: AUTHOR [+4]

Abstract

Described is a hardware implementation to provide a method whereby the users of speech recognition systems will be able to pace their input. The implementation involves the use of buttons and indicators to indicate to the user the various states to operate the system and assures recognition when required.

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

Method to Pace User Input for Speech Recognition Systems

      Described is a hardware implementation to provide a method
whereby the users of speech recognition systems will be able to pace
their input.  The implementation involves the use of buttons and
indicators to indicate to the user the various states to operate the
system and assures recognition when required.

      Typically, speech recognition systems stop listening for speech
at some point in the operation and are placed into a wait state.  The
stopping generally occurs as a result of a program generated
interrupt, a user command, a timer, or some other internal or
external stimulus.  There are many reasons for a speech server to be
in a wait state, rather than in a listen state, such as when an
application is idle, or an execution must complete a prior command
before continuing to the next application.

      In prior art, if a user attempted to perform operations, such
as speaking into a speech recognition device, the system might not be
ready to accept the input and an error would be generated.  Simply
indicating that the microphone was off proved to be insufficient
since users were unable to tell if the microphone needed to be
manually turned on, or was temporarily turned off, such as during a
screen refresh or a data retrieval operation.

      The concept described herein provides a multi-state actuating
device to indicate the status of the system's listening mode and
actuating buttons to change states.  The multi-state actuating
devices consist of three buttons and three indicators that provide
the state changes initiated under program control.  The three states
are as follows:
 1.  Talk State - In this state, the speech recognition system is
    waiting, or listening.  An end-of-utterance signal causes the
    system to provide a transition to the wait state and program
    intervention causes the system to transition to a final, or off
    state, such as a time-out.  The user can turn the microphone off,
    or the user can manually identify an end-of-utterance by
    releasing a push-to-talk button on a controller unit imbedded in
    the microphone.
 2.  Initial State - In this state, the microphone is temporarily
off.
    The system is waiting for some other operation other than a
    spoken input.  Usually, the system, or application, is performing
    other tasks.
 3.  Final State -...