Browse Prior Art Database

Speech Recognition in Public Access Kiosks

IP.com Disclosure Number: IPCOM000118776D
Original Publication Date: 1997-Jul-01
Included in the Prior Art Database: 2005-Apr-01
Document File: 2 page(s) / 48K

Publishing Venue

IBM

Related People

Byford, DJ: AUTHOR

Abstract

Some Speech-Recognition (S-R) products have reached levels of development and price-performance that means they can be considered for use with speaker independent applications such as public-access, multi-media kiosks. Speech recognition can be used in such kiosks for controlling the display by responding to menus, entering information on electronic forms, and requesting information. Speech-recognition systems for such general use set particular demands for accuracy and tolerance of a wide range of speaking styles (speaker independence).

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 64% of the total text.

Speech Recognition in Public Access Kiosks

      Some Speech-Recognition (S-R) products have reached levels of
development and price-performance that means they can be considered
for use with speaker independent applications such as public-access,
multi-media kiosks.  Speech recognition can be used in such kiosks
for controlling the display by responding to menus, entering
information on electronic forms, and requesting information.
Speech-recognition systems for such general use set particular
demands for accuracy and tolerance of a wide range of speaking styles
(speaker independence).

      The design and method described here maximizes the accuracy and
speaker independence of speech recognition in public-access kiosks.

      The improved kiosk incorporates three parallel implementations
of the speech recognition software, each one linked to a separate
microphone.  The microphones are positioned in optimal positions in
the acoustically designed hood of the kiosk.  Each version of the S-R
software has been 'pre-trained' by a different group of individuals
(each with a mix of accents but weighted towards the regional accent
used around the proposed location of the kiosk).  This ensures that
each version of the software is distinct and has a 'unique enrollment
experience' to draw on.  An additional piece of software (a
'consensus routine') processes the output of the three versions of
the S-R software  which will consist of the suggested word
recognized,...