Ex Parte II et al

Board of Patent Appeals and Interferences

Mar 23, 2012

10294534 (B.P.A.I. Mar. 23, 2012)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
10/294,534 11/14/2002 David L. Ii LM(F)5713 1371
26294 7590 03/26/2012
TAROLLI, SUNDHEIM, COVELL & TUMMINO L.L.P.
1300 EAST NINTH STREET, SUITE 1700
CLEVELAND, OH 44114
EXAMINER
CARTER, AARON W
ART UNIT PAPER NUMBER
2624
MAIL DATE DELIVERY MODE
03/26/2012 PAPER
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE BOARD OF PATENT APPEALS
AND INTERFERENCES
____________

Ex parte DAVID L. LI, ELLIOTT D. REITZ, II,
and DENNIS A. TILLOTSON
____________

Appeal 2009-013219
Application 10/294,534
Technology 2600
____________

Before JOSEPH F. RUGGIERO, MAHSHID D. SAADAT,
and CARL W. WHITEHEAD, JR., Administrative Patent Judges.

SAADAT, Administrative Patent Judge.

DECISION ON APPEAL
Appellants appeal under 35 U.S.C. § 134 from the final rejection of
claims 1-6 and 8-20. Claim 7 has been canceled and claim 21 has been
indicated as containing allowable subject matter. We have jurisdiction under
35 U.S.C. § 6(b).
We affirm.
Appeal 2009-013219
Application 10/294,534

2
STATEMENT OF THE CASE
Introduction
Appellants’ invention relates to methods and computer program
products for pattern recognition employing an optimal confidence threshold
value to maximize the accuracy of the classification of the received stimuli
(see Spec. 3:9 - 4:2).
Exemplary independent claim 1 reads as follows:
1. A method of determining an efficient set of
features and an optimal threshold confidence value for a pattern
recognition system with at least one output class, comprising:
selecting an initial set of features based upon an
optimization algorithm;
classifying a plurality of pattern samples using the
selected feature set;
optimizing a threshold confidence value to maximize the
accuracy of the classification;
accepting the selected feature set and threshold
confidence value if a cost function based upon classification
accuracy meets a predetermined threshold cost function value;
and
changing the feature set, by adding, removing or
replacing a feature within the set based upon the optimization
algorithm, if the cost function does not meet the predetermined
threshold cost function value.
Rejection
The Examiner rejected claim 1 under 35 U.S.C. § 102(b) as being
anticipated by Gaborski (US 5,479,523).12

1 Separate patentability was not argued with any specificity for the
remaining claims rejected under §§ 102 and 103 over Gaborski (App. Br. 7).
2 Throughout this opinion, we refer to the second or subsequent Appeal Brief
filed Feb. 12, 2009, and the second or subsequent Examiner’s Answer
mailed May 18, 2009.
Appeal 2009-013219
Application 10/294,534

3
Appellants’ Contentions
Appellants contend that Gaborski does not support a finding of
anticipation because the cited portions of the reference fail to disclose an
optimal threshold confidence value for a classifier represented by a given
feature set (App. Br. 5). Appellants specifically assert:
[b]ut the classification efficiency of Gaborski is simply an
optimization value used as part of the processing of selecting a
set of features, comparable to the cost function recited in claims
1 and 11. There is no teaching of a second optimization process
that determines a threshold confidence value for each set of
features, as recited in claims 1 and 11.
(App. Br. 6).
Gaborski is not clear as to how low confidence classifications
are handled by a classifier system trained by the methodology
taught by Gaborski, and a system utilizing a preselected
threshold confidence value or no threshold confidence value
would be operative, especially in an environment in which the
number of output classes likely to be encountered by the system
is limited, such as the optical character recognition system
described in Gaborski.
(Id.).

Appellants further contend that the maximum classification efficiency
of Gaborski, which is characterized by the Examiner as the claimed
optimizing a threshold confidence value (see Ans. 8), does not comprise
selecting a threshold confidence value for a classification for a set of
features, as recited in claim 1 (Reply Br. 4).

ANALYSIS
We agree with the Examiner and adopt as our own (1) the findings
and reasons set forth by the Examiner in the action from which this appeal is
Appeal 2009-013219
Application 10/294,534

4
taken and (2) the arguments set forth by the Examiner in the Examiner’s
Answer in response to Appellant’s Appeal Brief (see Ans. 12-18).
For emphasis, we note that Gaborski discloses the claimed
“optimizing a threshold confidence value to maximize the accuracy of the
classification” as determining a maximum “classification efficiency” (see
col. 5, ll. 6-13). Gaborski further discloses that the classification weight
matrices are “easy to manipulate by virtue of their being formed utilizing the
aforementioned, reduced size, feature subset having a maximum
classification efficiency” (col. 5, ll. 14-21). Therefore, contrary to
Appellants’ argument that Gaborski fails to teach the disputed claim feature
(Reply Br. 4), Gaborski ’s maximum classification efficiency is also
optimized when the space of f-elements is searched for an f-element subset
having a maximum classification efficiency (see Gaborski; col. 5, ll. 38-56).
In other words, selecting a different subset optimizes the threshold
confidence value when a subset with the maximum classification efficiency
is searched for and selected.
We also note that Appellants’ argument alleging the absence of
“selecting a threshold confidence value for a classification for a set of
features” in the disclosure of Gaborski is not convincing of error in the
Examiner’s rejection because the claim merely requires “optimizing a
threshold confidence value,” and not selecting such a value for a selected set
of features. In that regard, Gaborski discloses “optimizing a threshold
confidence value to maximize the accuracy of the classification” when a
subset of f-elements having a maximum classification efficiency is selected
(see Gaborski; col. 5, ll. 49-56).

Appeal 2009-013219
Application 10/294,534

5
CONCLUSION
On the record before us, we conclude that, because Gaborski teaches
all the claim limitations, the Examiner has not erred in rejecting claim 1 as
being anticipated by Gaborski. Therefore, we sustain the rejections of
claims 1-6, 8, 9, and 11-19 under 35 U.S.C. § 102(b) and of claims 10 and
20 under 35 U.S.C. § 103(a).

DECISION3
The Examiner’s decision rejecting claims 1-6 and 8-20 is affirmed.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(1)(iv).

AFFIRMED

ke

3 We have decided the appeal before us. However, we observe that claim 11
recites “a computer program product, implemented on a computer readable
medium” without specifying that such program is stored on the medium or is
in a “non-transitory” state. In case of further prosecution of claims 11-20,
the Examiner’s attention is directed to recently issued guidance from the
Director and our reviewing courts. See In re Nuijten, 500 F.3d 1346 (Fed.
Cir. 2007), and Subject Matter Eligibility of Computer Readable Media,
1351 OFFICIAL GAZETTE U.S. PAT. & TRADEMARK OFF. 212 (Feb. 23, 2010).