Appeal No. 1996-2198
Application 08/077,506
(3) C, which defines that the probability that the
sequence of events will begin at state i (Brief at 10).
Page 7 of the brief shows that the state transition
probability distribution A can be represented as a matrix of
discrete probability values. As will appear, it was known in
the art to represent the occurrence probability distribution B
as either (1) a set of discrete probability values obtained
from quantized feature vectors or (2) as a continuous
probability function derived from non-quantized feature
vectors. The § 103 question before us is whether it would
have been obvious to represent the occurrence probability
distribution as an approximate continuous probability function
Bc derived from a set of discrete probability values B
obtained from quantized feature vectors.2
2The Answer was accompanied by a copy of Rabiner &
Juang, "An Introduction to Hidden Markov Models," IEEE ASSP
Magazine, January 1986, pp. 12-15, which the examiner cites as
"teach[ing] that it was obvious to extend any discrete model
by substituting [sic, replacing] discrete probability
functions with continuous density functions" (Answer at 5).
This publication will not be considered, because it is not
mentioned in the statement of the § 103 rejection and was
cited for the first time in the answer. See Ex parte Movva,
(continued...)
-3-
Page: Previous 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Next
Last modified: November 3, 2007