Appeal No. 1997-4121
Application No. 08/427,272
Rather, it is the combination of the word spotting scheme and the specific learning scheme which
distinguishes over the applied prior art.
While word spotting was well known, the references to Takebayashi and Bahler are alleged by
appellants to be deficient in showing any word spotting schemes. Since word spotting is a continuous
pattern matching method in which the word boundary is taken to be not fixed but flexible (see the definition
at lines 3-4 of page 2 of the specification), it does appear that neither Takebayashi nor Bahler discloses
such a method. Takebayashi uses a recognition means which detects starting and ending points, thus fixing
the word boundary. While Bahler does disclose the recognition of keywords in a “continuous” audio
signal, and would thus appear to be indicative of a word boundary which is not fixed, further review of the
reference would appear to indicate otherwise. For example, the reference discloses a plurality of
templates representing silence, or nonspeech, signals, which would appear to indicate that these templates
would set some kind of boundary around a spoken word. In any event, appellants submit, as evidence, a
declaration under 37 CFR 1.132, from Mr. Hidenori Shinoda, a clearly skilled artisan. Mr. Shinoda
explains that the concatenation technique of Bahler “is a technique for connecting two or more specific
word templates in order to detect word sequences..., which is part of the syntax control used in
determining beginning and ending boundaries of unknown keywords in Bahler.” Mr. Shinoda explains,
further, that this “technique is totally unrelated to whether the recognition is ‘word spotting’ or not, because
the difference between ‘word spotting’ and ‘non-word spotting’ is not a matter of how word templates are
5
Page: Previous 1 2 3 4 5 6 7 8 9 Next
Last modified: November 3, 2007