Browse Prior Art Database

Speech Recognition Enhancement Utilizing Statistical Predication of Letter Sequence

IP.com Disclosure Number: IPCOM000114501D
Original Publication Date: 1994-Dec-01
Included in the Prior Art Database: 2005-Mar-28
Document File: 2 page(s) / 44K

Publishing Venue

IBM

Related People

Osborn, NA: AUTHOR

Abstract

Described is a method whereby the utilization of speech recognition, as used in computer technology, is enhanced by using spoken letter recognition techniques. The concept postulates the use of a search criteria by utilizing the statistical prediction of a letter sequence based on the probability of one letter following another.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 67% of the total text.

Speech Recognition Enhancement Utilizing Statistical Predication
of Letter Sequence

      Described is a method whereby the utilization of speech
recognition, as used in computer technology, is enhanced by using
spoken letter recognition techniques.  The concept postulates the use
of a search criteria by utilizing the statistical prediction of a
letter sequence based on the probability of one letter following
another.

      Typically, speech recognition requires a great deal of computer
system resources, both in terms of specialized hardware, processing
power and storage of the recognized vocabulary.  Difficulty arises in
supporting these requirements in small hand-held computer devices.

      In prior art, the incorporation of a speech interface for
computer activation was to limit the size of the recognized
vocabulary.  With a limited vocabulary, it was desirable to include
the names of the letters, such as A or B, etc., within a set in order
to allow the flexibility to spell words that were not contained
within the limited recognition vocabulary.  However, the individual
spoken letters were difficult to identify.  The sequence of sounds
that made up a letter name was generally short.  Also, many letter
names have similar sounds, such as B, D and V.  Machine speech
recognition techniques have attempted to prioritize the selection of
phonemes by utilizing the probability of a phoneme following one that
has already been recognized.  Homonyms were resol...