Appeal No. 95-4221 Page 2
Application No. 08/086,569
language, but the specification identifies no corresponding
structure, so they are principally defined by their function.
Consequently, we select method claim 6 as representative of
all of the claims on appeal. The examiner has rejected
claims 4-8 as anticipated under 35 U.S.C. § 102(e) by
Bahl et al. (Bahl) 5,033,087 16 July 1991
filed 14 Mar. 1989
We reproduce claim 6 below with enumeration and labels from
Bahl indicating the examiner's reading of the claim on the
reference:
A continuous speech recognition method,
comprising the steps of:
(1) in response to a currently verified phoneme,
predicting a subsequent phoneme using an action
entry in a stored left to right (LR) parser
table 1030, (8:1-31 and 53-54);
(2) predicting a phoneme context for the
predicted subsequent phoneme;
(3) verifying existence of the predicted
subsequent phoneme in the input speech signal (phone
machines, 6:48-62) using a phoneme context dependent
type hidden Markov phoneme model (Figs. 4A & 4B
Markov models) which corresponds to the predicted
phoneme context to calculate a probability that the
predicted subsequent phoneme exists in the input
speech signal;
(4) executing steps (1) through (3) repeatedly,
each repetition using the predicted subsequent
phoneme as a new currently verified phoneme to
thereby produce a symbol string of verified phonemes
Page: Previous 1 2 3 4 5 Next
Last modified: November 3, 2007