Ex Parte Cobb et al

Patent Trial and Appeal Board

Sep 19, 2016

12543141 (P.T.A.B. Sep. 19, 2016)

UNITED STA TES p A TENT AND TRADEMARK OFFICE
APPLICATION NO. FILING DATE
12/543,141 08/18/2009
110344 7590 09/21/2016
Patterson & Sheridan, LLP/BRS Labs
24 Greenway Plaza, Suite 1600
Houston, TX 77046
FIRST NAMED INVENTOR
Wesley Kenneth Cobb
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
ATTORNEY DOCKET NO. CONFIRMATION NO.
BRS/0021 6753
EXAMINER
CHU, RANDOLPH I
ART UNIT PAPER NUMBER
2666
NOTIFICATION DATE DELIVERY MODE
09/21/2016 ELECTRONIC
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.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address( es):
PSDocketing@pattersonsheridan.com
Pair_eOfficeAction@pattersonsheridan.com
VKubitskey@pattersonsheridan.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte WESLEY KENNETH COBB,
RAJKIRAN KUMAR GOTTUMUKKAL,
KISHOR ADINATH SAITWAL, MIN-JUNG SEOW,
GANG XU, LON WILLIAM RISINGER, and JEFF GRAHAM
Appeal2015-001938
Application 12/543,141
Technology Center 2600
Before JOSEPH L. DIXON, ERIC S. FRAHM, and
KRISTEN L. DROESCH, Administrative Patent Judges.
DIXON, Administrative Patent Judge.
DECISION ON APPEAL
Appeal2015-001938
Application 12/543,141
STATEMENT OF THE CASE
Appellants appeal under 35 U.S.C. § 134(a) from a rejection of claims
1-25. Claims 26-32 have been withdrawn from consideration. We have
jurisdiction under 35 U.S.C. § 6(b).
We affirm.
The claims are directed to a pixel-level based micro-feature
extraction. Claim 1, reproduced below, is illustrative of the claimed subject
matter:
1. A computer-implemented method for extracting pixel-
level micro-features from image data captured by a video
camera, the method comprising:
receiving the image data;
identifying a foreground patch that depicts a foreground
object;
processing the foreground patch to compute a plurality of
micro-feature values based on at least one pixel-level
characteristic of the foreground patch, wherein the micro-feature
value is computed independent of training data that defines a
plurality of object types;
generating a micro-feature vector that includes the
plurality of micro-feature values; and
classifying the foreground object as depicting an object
type as based on the micro-feature vector.
2
Appeal2015-001938
Application 12/543,141
REFERENCES
The prior art relied upon by the Examiner in rejecting the claims on
appeal is:
Boregowda et al.
Farmer et al.
Saptharishi et al.
US 2007 /0058836 Al
US 2008/0131004 Al
US 2012/0274777 Al
REJECTIONS
The Examiner made the following rejections:
Mar. 15, 2007
June 5, 2008
Nov. 1, 2012
Claims 1-3 and 5-25 stand rejected under 35 U.S.C. § 102(b) as being
anticipated by Boregowda. 1
Claim 4 stands rejected under 35 U.S.C. § 103(a) as being
unpatentable over Boregowda in view of Farmer.
Claims 4, 5, 14, and 15 stand rejected under 35 U.S.C. § 103(a) as
being unpatentable over Boregowda in view of Saptharishi.
ANALYSIS
With respect to independent claims 1, 13, and 18, Appellants argue
the claims together. (App. Br. 8). Therefore, we select independent claim 1
1 We note that the Examiner appears to reject dependent claim 6 based upon
obviousness and does not include a specific paragraph addressing claim 6
with regards to the anticipation rejection, but includes a specific paragraph
regarding the rejection based upon the combination of Boregowda in view of
Saptharishi. (Final Act. 9). Additionally, we note that the Examiner
mentions claim 4 in two rejections in the Final Action. First, The Examiner
mentions claim 4 in the rejection based upon obvious over Boregowda alone
and in the introductory paragraph of the combination of Boregowda in view
of Saptharishi. (Final Act. 8 and 9). We interpret claim 4 to be rejected
based upon Boregowda alone and claim 6 to be rejected based upon
obviousness over the combination of Boregowda in view of Saptharishi.
3
Appeal2015-001938
Application 12/543,141
as the representative claim for the group and will address Appellants'
arguments thereto.
We have reviewed the Examiner's rejections (Final Act. 4--5) in light
of Appellants' contentions in the Appeal Brief (App. Br. 8-11) and the
Reply Brief (Reply Br. 2--4) that the Examiner has erred, as well as the
Examiner's response (Ans. 4--9) to Appellant's arguments in the Appeal
Brief. We disagree with Appellants' conclusions.
We concur with the conclusions reached by 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 taken (Final Act. 4--5), and (2) the reasons
set forth by the Examiner in the Examiner's Answer in response to
Appellants' Appeal Brief (Ans. 4--9). We highlight and amplify certain
teachings and suggestions of the references, as well as certain ones of
Appellants' arguments as follows.
With respect to representative independent claim 1, Appellants argue
that the Examiner's reliance upon the Boregowda reference is in error
regarding the claimed "processing the foreground patch to compute a
plurality of micro-feature values based on at least one pixel-level
characteristic of the foreground patch, wherein the micro-feature value is
computed independent of training data that defines a plurality of object
types." (App. Br. 8). Appellants further argue:
As described in Boregowda, a foreground object may be
evaluated to determine whether that foreground object represents
an instance of a predefined class. However, this description of a
trained classifier in the present specification is expressly
distinguished from the claimed technique for using a classifier to
learn arbitrary classification types by clustering micro-feature
4
Appeal2015-001938
Application 12/543,141
vectors generated for foreground objects depicted in a video
stream.
(App. Br. 9). Appellants contend:
fundamental features disclosed in Boregowda are shape-
and segment-related features of a pixel blob that have length,
width, and height properties. This differs from "micro-feature
values based on at least one pixel-level characteristic of the
foreground patch." That is, micro-feature values are pixel-level
based properties in a pixel region that are used to generically
classify multiple object types.
(App. Br. 10).
The Examiner maintains:
In Boregowda, fundamental features (micro-feature
values) such as (length and width) are computed based on input
binary blob (the extreme white pixels in the MBR of blob) and
Shape features (micro-feature vector) such as circularity 242,
convexity 244, elongation indent 248 are computed based on
fundamental features, [ t ]hen ranges that the features may fall into
for a class (human, vehicle, etc.) are defined. So, fundamental
features are interpret[ ed] as micro-feature values and Shape
features are interpret[ed] as micro-feature vectors.
(Ans. 6). The Examiner further maintains:
In paragraph [0054] of Boregowda teach that "The derived
(generated) features such as the circularity, convexity, elongation
indent, and the projection histogram features (micro feature
vector) are computed for the second level of classification of the
blob. In this second classification, ranges that the features may
fall into for a class (human, vehicle, etc.) are defined, and class
weights are derived based on overlap made by feature ranges for
the different classes." "The derived features such as the
circularity, convexity, elongation indent, and the projection
histogram features 225" interprets as micro-feature vector and
the class (human, vehicle, etc.) are determined based on ranges
of the feature value.
5
Appeal2015-001938
Application 12/543,141
(Ans. 6-7). Additionally, the Examiner maintains:
As described in B, fundamental features (micro-feature
values) are computed based on input binary blob, then The shape
features (micro feature vector) are generated from fundamental
features (see Fig. 2), then the object is classified based on The
shape features (see A).
Boregowda teaches classify[ing] objects (blob) based on
generating a micro feature vector (the derived feature) and
determining an object type (such as human, vehicle and etc.)
[]based on the micro-feature vector (the derived feature).
Therefore, Boregowda does teach "computing a plurality
of micro-feature values (length, width) based on at least one pixel
characteristic (the extreme white pixels in the MBR) of the
foreground patch (blob).
(Ans. 7). The Examiner further disagrees with Appellants and maintains:
In Boregowda, predefined range of class such as human,
vehicle, other (para. [0065]) can be interpreted as training data
because the derived features from the blob are validated with
respect to the range of value of Predefined class to decide on the
class to which the blob belongs. In order to calculate the micro-
feature value (length and width), it only requires information
from the extreme white pixels in the MBR. It does not require
any information from predefined class such as value range of
human, vehicle. Therefore Boregowda does teach "wherein the
micro-feature value is computed independent of training data that
defines a plurality of object types."
It appears that nowhere in the claim describes that what is
"training data" in the claim or how it is related to micro feature
value. It can be understood as any random data that defines
object type such as metadata associated with the object. It can
be seen from Fig. 5 of Boregowda that when the system classifies
object 521 as vehicle, it does not use object 511 (human)
(training data) in order to classify the vehicle. When micro-
6
Appeal2015-001938
Application 12/543,141
feature value (length and width) is calculated, any information
from object 511 or 521 not needed. Therefore, it can be
concluded that the computation for "a vehicle", in this instance,
is independent from "training data" that defines other objects as
human.
(Ans. 7-8). The Examiner further maintains:
Boregowda teach that the derived features from the blob 205 are
validated if derived feature value lies in the predefined range of
the class which are human 440, vehicle 410, multiple human 430,
and other (420) ranges as illustrated by example in FIG. 4. (para.
[0065]). Which mean[ s] that features are derived from the blob,
not predefined classes. The derived features of Boregowda are
length and width. In order to compute length and width,
locations of two pixels are needed. Locations of two pixels can
be retrieved from the blob. In Boregowda, length and width
computed based on the extreme white pixels in the MBR
(minimum bounding rectangle) wherein white pixels are pixels
around the perimeter of the binary image (blob). There are four
extreme white pixels in the MBR, (Xmin, Y min), (Xmin, Y max),
(Xmax, Y min), (Xmax, Y max). Length can be calculated by Xmax -
Xmin and width can be calculated by Y max - Y min. Everything is
required to compute length and width can be retrieved from
extreme white pixels in the MBR of blob. Therefore, information
from predefined classes is not required to calculate length and
width.
(Ans. 8-9). We agree with the Examiner's findings of fact and the finding
of anticipation of independent claim 1.
Additionally, Appellants contend:
The fundamental features are values based on length,
width, and height features in an input blob. In contrast, micro-
feature values are pixel-level-based properties in a pixel region
that are used to generically classify multiple object types. That
is, such properties are based on pixel-level characteristics that are
later included in a micro-feature vector representing the
characteristics.
7
Appeal2015-001938
Application 12/543,141
(Reply Br. 3). Appellants further contend:
As described, length, width, and height values are
evaluated based on a collection of pixels that form an input blob.
That is, the "fundamental features" of Boregowda (e.g., length,
width, and height) are not calculated per pixel, but rather based
on the entirety of the input blob itself. In contrast, the micro-
feature values are computed based on characteristics on a pixel-
level. Clearly, the fundamental features disclosed in Boregowda
differ from the micro-features based on at least one-pixel level
characteristic of a foreground patch.
(Reply Br. 4).
We find Appellants' arguments are not commensurate in scope with
the express language recited in the language of independent claim 1. We
find a per-pixel calculation is not recited in language of independent claim 1.
Moreover, from our review of Appellants' disclosed invention,
Appellants' Specification evidences that the "micro-feature extractor and
micro-classifier may be included within a behavior-recognition system ...
In a particular embodiment, the behavior-recognition system may include
both a computer vision engine and a machine learning engine." (Spec. i-f 33).
Additionally, Appellants' Specification sets forth:
Data output from the computer vision engine may be
supplied to the machine learning engine. In one embodiment, the
machine learning engine may evaluate the context events to
generate "primitive events" describing object behavior. Each
primitive event may provide some semantic meaning to a group
of one or more context events. For example, assume a camera
records a car entering a scene, and that the car turns and parks in
a parking spot. In such a case, the computer vision engine could
initially recognize the car as a foreground object; classify it as
being a vehicle, and output kinematic data describing the
position, movement, speed, etc., of the car in the context event
stream. In tum, a primitive event detector could generate a
8
Appeal2015-001938
Application 12/543,141
stream of primitive events from the context event stream such as
"vehicle appears," ["]vehicle turns," "vehicle slowing," and
"vehicle stops" (once the kinematic information about the car
indicated a speed of 0). As events occur, and re-occur, the
machine learning engine may create, encode, store, retrieve, and
reinforce patterns representing the events observed to have
occurred, e.g., long-term memories representing a higher-level
abstraction of a car parking in the scene - generated from the
primitive events underlying the higher-level abstraction. Further
still, patterns representing an event of interest may result in alerts
passed to users of the behavioral recognition system.
(Spec. i-f 35). Appellants' Specification further sets forth:
The computer vision engine 13 5 may be configured to
analyze this raw information to identify foreground patches
depicting active objects in the video stream, extract micro-
features, and derive a variety of metadata regarding the actions
and interactions of such objects, and supply this information to a
machine learning engine 140. In tum, the machine learning
engine 140 may be configured to classify the objects, evaluate,
observe, learn and remember details regarding events (and types
of events) that transpire within the scene over time.
(Spec. i-f 41). We further find paragraphs 33, 35, and 41 of Appellants'
Specification evidences the disclosed invention is a learning-based system
which does not initially start with any specific training data. We find after
the system learns and stores data, the data learned from is training data up to
that point in time. We note that the language of independent claim 1 does
not set forth any details regarding the behavior recognition system, the
vision system, or learning system to distinguish the claimed invention from
the Boregowda reference. As a result, Appellants' arguments do not show
error in the Examiner's finding of anticipation of independent claim 1.
Appellants have not presented separate arguments for patentability of
9
Appeal2015-001938
Application 12/543,141
independent claims 13 and 18. Consequently, we group these claims as
falling with representative independent claim 1.
Appellants do not set forth separate arguments for patentability of
claims 2-12, 14--17, and 19-25. (App. Br. 11). As a result, these claims
will fall with their respective parent claims. The Examiner has not set forth
separate arguments for patentability regarding the obviousness rejections.
As a result, the Examiner has not shown error in the Examiner's conclusion
of obviousness.
CONCLUSION
The Examiner did not err in finding representative independent claim
1 to be anticipated by Boregowda. Because the Examiner has not set forth
separate arguments for patentability regarding the obviousness rejections of
claims 4, 5, 14, and 15, we sustain the obviousness rejections for the same
reasons as representative claim 1.
DECISION
For the above reasons, we sustain the Examiner's rejections of claims
1-25.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(l)(iv).
AFFIRMED
10