Browse Prior Art Database

Automatic Creation of Multiple Speech Recognition Voice Profiles

IP.com Disclosure Number: IPCOM000114637D
Original Publication Date: 1995-Jan-01
Included in the Prior Art Database: 2005-Mar-29
Document File: 2 page(s) / 44K

Publishing Venue

IBM

Related People

Barcello, M: AUTHOR

Abstract

Described is a method to automatically create multiple speech profiles in speech recognition, as used in computer systems. The technique provides a means of improving speech recognition functions through the use of multiple profiles, where one profile is created for each sound environment. This requires the user to train the system multiple times.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 73% of the total text.

Automatic Creation of Multiple Speech Recognition Voice Profiles

      Described is a method to automatically create multiple speech
profiles in speech recognition, as used in computer systems.  The
technique provides a means of improving speech recognition functions
through the use of multiple profiles, where one profile is created
for each sound environment.  This requires the user to train the
system multiple times.

      Generally, computer systems designed for speech recognition
purposes require that the person intending to use the system first go
through an enrollment process.  The enrollment process typically
involves reading words, or sentences, into a microphone.  Information
is then processed through the computer's recognition system to create
a unique voice profile for the enrolled user.  This profile is then
used by the system to improve the accuracy of voice recognition when
the user is dictating.

      However, the accuracy of the voice recognition system can be
affected by the ambient acoustics and sound.  If a user trains the
system in a quiet environment, but uses the system in a noisy
environment, the accuracy of the voice recognition can be impaired.
The process of creating the profiles can consume a considerable
amount of time, typically one to two hours per enrollment for the
average dictation type of equipment.  This in itself can reduce
acceptance of speech recognition systems.

      To overcome this limitation, the concept d...