Browse Prior Art Database

Example Enrollment Text Playback for an Automatic Speech Recognizer

IP.com Disclosure Number: IPCOM000104188D
Original Publication Date: 1993-Mar-01
Included in the Prior Art Database: 2005-Mar-18
Document File: 2 page(s) / 50K

Publishing Venue

IBM

Related People

Daggett, G: AUTHOR [+4]

Abstract

Disclosed is a method to aid a user in properly pronouncing an enrollment script for a speech recognition system and in speaking in the correct manner during the enrollment process.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 54% of the total text.

Example Enrollment Text Playback for an Automatic Speech Recognizer

      Disclosed is a method to aid a user in properly  pronouncing an
enrollment script for a speech recognition system and in speaking  in
the  correct  manner  during  the   enrollment process.

      Speaker-dependent  automatic  speech  recognition (ASR) systems
require an enrollment process during which a user records a voice
sample.  These samples are read from a predetermined script of words
or phrases which the ASR enrollment system uses to generate models of
the user's speech patterns.  These models are later used to allow
recognition of words which the user speaks.

      The enrollment process expects a user to speak the script in a
manner consistent with the pronunciation of the words and phrases in
that script.  If the script is not spoken correctly, the speech
models of the user's voice will not be properly generated, and later
recognition of the user's voice will not be accurate.

Two types of errors are common when dictating an enrollment script:

o   Mispronouncing  a  word  or  phrase in the script, often because
    the word has an  ambiguous  pronunciation  (e.g., "read") or
    because it is a technical term or acronym (e.g., "AAA").

o   Words are not spoken in the mode  expected  by  the  ASR system,
    for example, if the user does not speak with the required pauses
    between words to an isolated-word ASR system.

In either case, th...