Ex Parte Cobb et al

Patent Trial and Appeal Board

Mar 14, 2018

13839587 (P.T.A.B. Mar. 14, 2018)

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.
13/839,587 03/15/2013 Wesley Kenneth COBB OMAI-002/03US
330076-2089
3417
58249 7590 03/16/2018
COOLEY LLP
ATTN: Patent Group
1299 Pennsylvania Avenue, NW
Suite 700
Washington, DC 20004
EXAMINER
BILLAH, MASUM
ART UNIT PAPER NUMBER
2486
NOTIFICATION DATE DELIVERY MODE
03/16/2018 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):
zpatdcdocketing@cooley.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte WESLEY KENNETH COBB, MING-JUNG SEOW,
GANG XU, KISHOR ADINATH, ANTHONY AKINS,
KERRY JOSEPH, and DENNIS G. URECH
Appeal 2017-008305
Application 13/839,587
Technology Center 2400
Before DENISE M. POTHIER, JUSTIN BUSCH, and JASON M. REPKO,
Administrative Patent Judges.
BUSCH, Administrative Patent Judge.
DECISION ON APPEAL
Pursuant to 35 U.S.C. § 134(a), Appellants appeal from the
Examiner’s decision to reject claims 1—21, which constitute all the claims
pending in this application. We have jurisdiction over the pending claims
under 35 U.S.C. § 6(b).
We reverse.
CLAIMED SUBJECT MATTER
Appellants’ invention generally “relate[s] to configuring a behavioral
recognition-based video surveillance system to generate alerts for certain
events.” Spec. 12. More specifically, Appellants’ invention allows “a
behavioral recognition system to identify events that should always or never
Appeal 2017-008305
Application 13/839,587
result in an alert without impeding the unsupervised learning process of the
surveillance system.” Id. Appellants’ claimed system and methods are
directed to aspects of a behavioral recognition video surveillance system,
which is a system that learns objects and behaviors over time rather than
relying on predefined rules and patterns. Id. 3^4. Appellants’ invention
provides the ability to override learned alert behavior—i.e., to specify
certain behaviors that should always or never result in an alert. Id. Tflf 5—6.
Claims 1, 8, and 15 are independent claims. Claim 1 is illustrative and
reproduced below:
1. A method for processing events generated from an
evaluation of a stream of video frames, the method comprising:
obtaining characteristic values for an observed event in a
scene depicted by the stream of video frames;
updating a learned state of the scene based on the
characteristic values, wherein the learned state provides a model
of patterns of behavior generated from evaluating a plurality of
foreground objects detected in the stream of video frames;
parsing a list of alert directives for a matching alert
directive having ranges of criteria values, wherein the
characteristic values are within the ranges of the criteria values
and wherein the alert directive overrides a decision to either
publish an alert or to not publish an alert for the observed event
based on the updated learned state of the scene without changing
the updated learned state of the scene; and
upon identifying the matching alert directive, either
publishing the alert or not publishing the alert according to alert
directive.
REJECTION
Claims 1—21 stand rejected under AIA 35 U.S.C. § 103(a) as obvious
in view of Venetianer (US 2005/0162515 Al; July 28, 2005) and Xu (US
2008/0181453 Al; July 31, 2008). Final Act. 2-11.
2
Appeal 2017-008305
Application 13/839,587
ANALYSIS
The Examiner rejects claims 1—21 as obvious in view of the combined
teachings of Venetianer and Xu. Final Act. 2—11. Of particular note, the
Examiner finds Venetianer does not disclose the updating and parsing steps,
but finds Xu teaches or suggests the updating and parsing steps. Id. at 4—5.
The Examiner provides a rationale for combining the relevant teachings of
Venetianer and Xu, concluding the claimed subject matter would have been
obvious in view of the proposed combination. Id. at 6.
Venetianer is generally directed to a video surveillance system that
uses event discriminators to extract event occurrences from video primitives
and generate a response, such as an alarm, based on the event occurrences.
Venetianer, Abstract. Venetianer allows an operator monitoring a location
to enter event discriminators based on video primitives that may be
evaluated alone or with other event discriminators to determine whether to
trigger a response, such as activating alerts, locking a door, and activating
another surveillance system. Id. Tflf 75—76, 97—98, 120, 137.
Xu is directed to a video surveillance system for tracking objects
using appearance models for blobs (a group of pixels that may be associated
with a particular object) to maintain object identity during an occlusion
event (when the two blobs overlap). Xu, Abstract. Xu uses the appearance
models to segment a group blob resulting from an occlusion event into
regions that may be identified with the particular blobs involved in the
occlusion event. Id.
Xu depicts and describes the stages “of a known intelligent video
system.” Xu 125, Fig. 1. Specifically, the disclosed known system includes
nine stages, grouped into three main blocks: object segmentation, robust
3
Appeal 2017-008305
Application 13/839,587
tracking, and object classification. Id. 13, Fig. 1. Xu’s system replaces the
known robust tracking block with its own matching process. Id. 39, 49,
Fig. 3.
In the first stage of the object segmentation block, the system learns a
background model based on an initial video segment. Xu 14. “The purpose
of this stage 7 is to establish a background model from an initial segment of
video data” and the background model ideally has no visible foreground
objects. Id. 141. Xu’s second stage of object segmentation is background
subtraction, which compares each pixel of a current frame to the background
model to estimate whether the pixel represents the background or a
foreground object and may dynamically update small changes in the
background model (whereas “more severe or sudden changes may require a
relearning operation”). Id. 142. Next, Xu implements a false-foreground
suppression stage, “attempts to alleviate false detection problems caused by
noise and camera jitter,” followed by a shadow/highlight removal stage and
the constrained component analysis (CCA) stage, which is the final stage of
object segmentation and “groups all pixels presumably belonging to
individual objects into respective blobs.” Id. Tflf 45—48. Xu distinguishes
between blobs, which it describes as newly detected foreground elements or
regions, and objects, which are foreground elements or regions that are being
tracked. Id. 149.
Xu’s matching process begins by analyzing each blob and assigning it
one of four “attention levels” based on predefined rules. Xu 1 50.
Subsequent processing of each blob depends on the attention level assigned
to that blob. Id. The attention levels indicate the status of a blob with
respect to potential occlusion with another blob. More particularly, attention
4
Appeal 2017-008305
Application 13/839,587
level 1 is assigned to blobs that are not overlapping or close enough to the
nearest neighbor to be imminently approaching an occlusion event. Id.^ 51.
If a blob is not overlapping, but within a proximity threshold of a neighbor,
it is assigned attention level 2, whereas attention level 3 indicates occlusion
is taking place, and attention level 4 indicates occluded objects are
separating. Id. Tflf 52—61. When blobs are about to occlude (i.e., the blobs
are assigned attention level 2), Xu creates or updates an appearance model
that is used when blobs are assigned attention levels 3 and 4 to perform more
accurate blob detection and segmentation. Id. 82—86, 90—99.
Appellants contend Xu does not teach or suggest either the updating
step or the parsing step recited in the independent claims. Br. 9—11. In
particular, with respect to the parsing step, Appellants assert the cited
portion of Xu addresses a well-known video analytics problem of occluding
foreground objects, which Xu addresses by changing how appearance
models are updated to prevent one object from contaminating the other. Id.
at 11. Appellants argue the appearance model updating described by Xu
simply indicates sensitivity to interacting foreground objects but in no way
teaches or suggests “a step of overriding whether alert messages are
published for certain behavioral events observed for certain foreground
objects.” Id.
The Examiner responds that Appellants’ argument “fails to account
for the combination of Venetianer and Xu as a whole.” Ans. 4. The
Examiner finds Xu teaches analyzing incoming frames to identify blobs and
assign each blob an attention level that determines further processing steps.
Id. at 5 (citing Xu 150). The Examiner further finds Xu “teaches that match
decision stage 53 created or updated for the relevant blobs depends on
5
Appeal 2017-008305
Application 13/839,587
whether or not a match is made,” and Xu’s appearance model “comprises a
color histogram indicating the frequency (i.e. number of pixels) of each
color level that occurs within that blob.” Id. at 4—5. The Examiner then
states “[o]nce attention manager assigned one of four possible levels based
on predefined rules, it would be possible to publish or not publish an alert
with the help of current system by recalling those assigned levels.” Id. at 5.
Finally, the Examiner finds Venetianer discloses that real-time video
primitive extraction enables the system to generate real-time alerts. Id. at 6
(citing Venetianer || 97, 119, 130). Thus, the Examiner concludes, the
parsing step, and specifically the recited “wherein the alert directive
overrides a decision to either publish an alert or to not publish an alert for
the observed event based on the updated learned state of the scene without
changing the updated learned state of the scene,” would have been obvious
in view of Venetianer and Xu.
Appellants do not dispute that Venetianer teaches generating system
alerts or that Xu teaches creating or updating appearance models for blobs
assigned attention level 2. Rather, Appellants argue Xu’s attention levels
and appearance models, discussed in paragraph 82 and cited by the
Examiner, do not “teach, or in any way suggest” the recited alert directives
that override a decision on whether or not to publish an alert. Br. 11. We
agree with Appellants that the Examiner’s rejection fails to demonstrate or
explain sufficiently how the proposed combination renders the claimed
subject matter obvious. In particular, Xu’s appearance models are not
related to alerts, let alone overriding a decision on whether or not to publish
an alert.
6
Appeal 2017-008305
Application 13/839,587
The Examiner’s explanation that, after determining an attention level,
“it would be possible to publish or not publish an alert with the help of
current system by recalling those assigned levels” is similarly deficient.
Ans. 5. Even to the extent that it would be possible to publish or not publish
an alert based on the assigned attention levels, the Examiner has not
explained adequately a link between the assigned attention levels and an
alert, or how the assigned attention levels would suggest to an ordinarily
skilled artisan to override the decision to publish or not publish an alert
taught by Venetianer. See id. Xu’s attention levels are assigned and used to
indicate a blob’s relationship to occlusion events. Xu Tflf 50-61. Xu is
concerned with robustly dealing with occlusions and providing improved
segmentation of occluded blobs in order to better maintain individual object
identity. Id. 8—9, 123. Based on the record, the Examiner has not
demonstrated Venetianer and Xu teach or suggest alert directives that
override a decision to publish or not publish an alert. Nor has the Examiner
articulated a sufficient reason with a rational underpinning to modify the
combined teachings to arrive at Appellants’ claimed subject matter.
For the reasons discussed above, we are persuaded the Examiner erred
in finding the combination of Venetianer and Xu teaches or suggests the
“parsing” limitation, as recited in independent claims 1, 8, and 15. Because
our determination is dispositive of this appeal, we do not address
Appellants’ other arguments. Accordingly, we do not sustain the
Examiner’s rejection of independent claims 1, 8, and 15. For similar
reasons, we also do not sustain the Examiner’s rejections of claims 2—7,
9—14, and 16—21, which depend from claims 1, 8, and 15, respectively.
7
Appeal 2017-008305
Application 13/839,587
DECISION
We reverse the Examiner’s decision to reject claims 1—21 under
35 U.S.C. § 103(a).
REVERSED
8