Browse Prior Art Database

Speech-Recognition System Enrollment Program with Training Features

IP.com Disclosure Number: IPCOM000113815D
Original Publication Date: 1994-Oct-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 67K

Publishing Venue

IBM

Related People

Goldwyn, R: AUTHOR [+8]

Abstract

Disclosed is the use of an enrollment procedure, for a speech recognition system, as a tool for training the potential user to speak in a manner facilitating the speech recognition process. In the enrollment process, a potential user of the system speaks several sentences into the system, as visual indication is provided on the display screen whenever a word is not spoken in this manner.

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

Speech-Recognition System Enrollment Program with Training Features

      Disclosed is the use of an enrollment procedure, for a speech
recognition system, as a tool for training the potential user to
speak in a manner facilitating the speech recognition process.  In
the enrollment process, a potential user of the system speaks several
sentences into the system, as visual indication is provided on the
display screen whenever a word is not spoken in this manner.

      The enrollment procedure traditionally is based on the user
reading sentences displayed by the system, following the movement of
a "bouncing ball," which is time-driven regardless of the way the
user is speaking.  To enroll successfully, and to use the system
subsequently, a potential user must learn to use a speech pattern in
which each word is spoken as a discrete entity, preceded and followed
by a pause.  The enrollment process monitors the speaker, flagging
words not spoken in this way.

      Since learning is required to use this speech pattern, the
enrollment process is expanded to include training features.  The
system, knowing what the user is supposed to say, does enough
recognition to locate where the breaks between words should be as he
speaks.  For example, the system knows how many words should be
spoken in a particular sentence in the enrollment procedure.  If too
many words are spoken, the system informs the user that he must speak
the sentence again.  If too few words are spoken, either an automati...