Appeal No. 1997-3914
Application No. 08/384,457
one set for each of the existing speakers, as an intermediate
step in generating a single pattern. The concatenating for
this generation of a single pattern is considered as part of
Appellant’s analyzing step (b). Thus, Appellant’s argument
regarding step (b) in not persuasive.
In much the same fashion, Appellant argues that step (c)
of claim 9 is not met because step (b) is not met. (Brief-top
of page 7.)
However, since we have found Gillick to meet step (b) of
claim 9, it follows that Appellant’s argument regarding step
(c) is similarly unpersuasive.
We note that as a result of Appellant’s invention,
Appellant’s new speaker requires less data for speech
recognition than the data collected from each of the existing
speakers (last part of claim 9). This result is also met by
Gillick (as noted by the Examiner) wherein it states:
But, by using cluster spellings and cluster
models of the type described above, an end user can
train up a large vocabulary system by speaking a
relatively small percent of its vocabulary words.
(Column 13, lines 26-29.)
7
Page: Previous 1 2 3 4 5 6 7 8 9 10 11 Next
Last modified: November 3, 2007