Browse Prior Art Database

Enrollment Monitors for an Automatic Speech Recognizer

IP.com Disclosure Number: IPCOM000104801D
Original Publication Date: 1993-Jun-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 2 page(s) / 68K

Publishing Venue

IBM

Related People

Daggett, G: AUTHOR [+4]

Abstract

Disclosed are two methods to provide status monitoring information to users of speech recognition systems during the enrollment phase. These methods monitor the correct recording of speech samples and the progress of the enrollment process in terms of percent complete.

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

Enrollment Monitors for an Automatic Speech Recognizer

      Disclosed are two methods to provide status monitoring
information to users of speech recognition systems during the
enrollment phase.  These methods monitor the correct recording of
speech samples and the progress of the enrollment process in terms of
percent complete.

      Speaker-dependent Automatic Speech Recognition (ASR) systems
require an enrollment process during which a user records samples of
his or her voice as a prototype for later user during recognition.
If these samples are not properly recorded, either because of user
error or some malfunction of the recording system, then the ASR
system will not be able to later recognize the user's voice
accurately.  Some of the most common types of recording errors during
enrollment are due to users speaking too softly or loudly, or due to
some hardware or software malfunction in the recording system that
prevents the voice signal from registering in the recording.

      This problem can be solved by having an enrollment recording
monitor which can analyze the user's voice signal during recording.
Characteristics such as minimum signal volume and maximum signal
volume can be checked to make sure they are within limits acceptable
to the ASR system in which the user is enrolling.

      The recording monitor can be incorporated as part of the
enrollment process software which is being used for enrolling in the
ASR system.  In this way, monitoring of the recordings can be done as
the user is speaking, instead of during some later step in enrollment
processing.  When the recording monitor detects some error in the
voice signal, the user can be immediately notified so that he or she
can alter the volume of their voice or make the necessary corrections
to the recording apparatus.

   ...