Appeal No. 1996-2198
Application 08/077,506
error affects the accuracy of speech pattern
recognition. [Col. 5, lines 11-32.]
According to appellants, Juang avoids errors relating to
discrete representations by replacing the code vector table
and converting circuit of the Rabiner-Levinson system with a
function generator which generates sets of continuous
probability density functions (Brief at 13-14). Because the
§ 103 rejection is not based on Juang, we need not determine
whether this is an accurate characterization of Juang's
invention.
Appellants' invention, in contrast, does not discard the
code vector table (14, 14A) and converting circuit (15, 15A)
of the Rabiner-Levinson system. Instead, it adds a function
generator 17 which generates an approximate continuous
occurrence probability function Bc derived from the set of
discrete probability values B generated by the first pattern
forming circuit (16, 16A). This function can take the form of
either a Gaussian probability density function or a Poisson
probability density function (Spec. at 11, line 27 to p. 12,
line 6). The approximate continuous occurrence probability
-7-
Page: Previous 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Next
Last modified: November 3, 2007