Ex Parte RAPTIS et alDownload PDFPatent Trial and Appeal BoardSep 24, 201814025722 (P.T.A.B. Sep. 24, 2018) Copy Citation UNITED STA TES p A TENT AND TRADEMARK OFFICE APPLICATION NO. FILING DATE 14/025,722 09/12/2013 94472 7590 09/26/2018 DISNEY ENTERPRISES, INC. c/o Sheppard Mullin Richter & Hampton LLP 650 Town Center Drive, 4th Floor Costa Mesa, CA 92626 FIRST NAMED INVENTOR MICHAIL RAPTIS 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. 12-DIS-426-PR-US-UTL 2551 EXAMINER SHIBRU, HELEN ART UNIT PAPER NUMBER 2484 NOTIFICATION DATE DELIVERY MODE 09/26/2018 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): docketing@sheppardmullin.com SheppardMullin_Pair@firsttofile.com PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE PATENT TRIAL AND APPEAL BOARD Ex parte MICHAIL RAPTIS and LEONID SIGAL Appeal2018-001499 Application 14/025,722 Technology Center 2400 Before ELENI MANTIS MERCADER, JENNIFER L. McKEOWN, and JASON M. REPKO, Administrative Patent Judges. REPKO, Administrative Patent Judge. DECISION ON APPEAL STATEMENT OF THE CASE Appellants 1 appeal under 35 U.S.C. § 134(a) from the Examiner's rejection of claims 1-34. App. Br. 4. 2 We have jurisdiction under 35 U.S.C. § 6(b ). We reverse. 1 Appellants identify the real party in interest as Disney Enterprises, Inc., a wholly owned subsidiary of The Walt Disney Company. App. Br. 1. 2 Throughout this opinion, we refer to the Final Rejection ("Final Act.") mailed August 31, 2016, the Appeal Brief ("App. Br.") filed May 26, 2017, the Examiner's Answer ("Ans.") mailed October 5, 2017, and the Reply Brief ("Reply Br.") filed November 29, 2017. Appeal2018-001499 Application 14/025,722 THE INVENTION Appellants' invention relates to video-based action recognition. Spec. ,r,r 2-3. According to the Specification, most video-based action- recognition modeling computes features over long temporal trajectories with many video frames. Id. ,r 3. These approaches can be computationally expensive or sensitive to dropped frames. Id. To address these issues, Appellants' invention selects optimal video frames using encoded frames and correlations. Id. ,r 4. In particular, the invention uses poselets, which can be an image patch or a human body's three-dimensional configuration. Id. ,r 5. The method looks for an action-discriminative poselet in consecutive video frames. Id. ,r 28. According to the Specification, Appellants' method is computationally efficient because it uses parts of the modeled action and spends fewer resources on non-discriminative or irrelevant frames. Id. ,r 27. Claim 1 is reproduced below with our emphasis: 1. A computer-implemented method for recognizing actions, compnsmg: selecting a plurality of poselets that are components of an action; encoding each of a plurality of video frames as a summary of the detection confidence of each of the plurality of pose lets for the video frame; and encoding correlations between poselets in the encoded video frames. THE EVIDENCE The Examiner relies on the following as evidence: Belhumeur et al. US 2012/0308124 Al Dec. 6, 2012 2 Appeal2018-001499 Application 14/025,722 THE REJECTION Claims 1-34 stand rejected under 35 U.S.C. § 102(a)(2) as anticipated by Belhumeur et al. Final Act. 3-6. ANALYSIS The Examiner finds that Belhumeur discloses every recited element of representative claim 1. Final Act. 3--4. In particular, the Examiner finds that Belhumeur selects a plurality of poselets that are components of an action by detecting parts within an image of a human face. Id. at 4 ( citing Belhumeur ,r 37, claim 3). In the Examiner's view, Belhumeur's images are indicative of an action because they show a person smiling. Ans. 8. Furthermore, the Examiner finds that Belhumeur encodes video frames by aligning fiducial points, evaluating the fit, detecting image features, and computing a score. Final Act. 4 ( citing Belhumeur ,r,r 40--42, 48). According to the Examiner, Belhumeur encodes correlations by combining a global model and a score, and applying a best fit to the selected data. Final Act. 4 (citing Belhumeur ,r,r 43, 49). Appellants argue that Belhumeur does not process video frames. App. Br. 6. According to Appellants, Belhumeur processes a collection of photographs of human faces, and these photographs were not recorded by a video-recording device as frames in a video sequence. Id. at 7. That is, Appellants' arguments (id. at 6-7) raise the issue of whether Belhumeur discloses "encoding correlations between poselets in the encoded video frames," as recited in claim 1. We agree with Appellants that the Examiner erred in finding that Belhumeur discloses this step. 3 Appeal2018-001499 Application 14/025,722 The limitation at issue involves video frames and poselets. Poselets can be an image patch or a three-dimensional configuration of a human body. Spec. ,r 5. The poselets can capture a person's pose or an object's appearance that indicates action. Id. ,r 25. For example, a poselet may capture a bent leg, a ball, a player's head-shoulder outline, among other things. Id. One embodiment models an action's temporal component by encoding pairwise correlations between poselets in video frames. Id. ,r 41. The encoding involves determining whether a first poselet and a corresponding second poselet are present or absent in video-frame pairs. Id. This process creates a pairwise frame descriptor. Id. Although this embodiment is non-limiting, it nevertheless informs our understanding of the claimed subject matter. To address the recited correlations between poselets in the encoded video frames, the Examiner relies on Belhumeur's processing of photographs of faces. Belhumeur ,r,r 43, 49, cited in Final Act. 4. In the cited embodiment, Belhumeur detects image regions that contain a face and generates a score for locations in the image. Belhumeur ,r 42. This score represents the likelihood that the location contains a fiducial. Id. Next, Belhumeur generates a non-parametric global model and combines it with the scores. Id. ,r 43. Belhumeur's method then labels facial features found in the image. Id. ,r,r 44--45. Unlike the recited correlations involving video frames, Belhumeur operates on "a collection of face images acquired from internet search sites using simple text queries." Id. ,r 46. The Examiner notes that Belhumeur also uses a publicly available dataset containing "1,521 images, each 4 Appeal2018-001499 Application 14/025,722 showing a frontal view of a face of one of 23 different subjects." Ans. 11 (citing Belhmneur ,r 121). The Examiner has not shown that either dataset includes video. Final Act. 3--4. Rather, Belhumeur's description of the dataset indicates that the process operates on individual images. See Belhumeur ,r,r 46, 121. Notably, the Examiner-cited process is concerned with features within a single image, instead of claim 1 's video of an action or even an image sequence. Id. ,r,r 43--49. In fact, Belhumeur explains that the dataset used stands in contrast to those that use the same camera for all images, like a set of video frames. Id. ,r 46. The Examiner finds that smart phones, video-game consoles, digital cameras, as described in Belhumeur, are capable of capturing and processing video. See Ans. 6 ( citing Belhumeur ,r 40). Likewise, the Examiner asserts that Internet-search sites are also capable of providing video. Ans. 6. But in the rejection, the Examiner has not shown that Belhumeur uses video from any of these sources in the way claim 1 recites. Final Act. 3--4. On the weight of this evidence, the Examiner has not shown that Belhumeur discloses "encoding correlations between poselets in the encoded video frames." App. Br. 6-8. Although Appellants make other arguments, we find this argument dispositive in this appeal. See id. at 4--12; Reply Br. 3-8. Accordingly, we do not sustain the Examiner's rejection of independent claim 1 and independent claim 18, which also recites the limitation at issue and is rejected under the same rationale (Final Act. 6; Ans. 5). For similar reasons, we also do not sustain the rejection of claims 2-17 and 19-34, which depend from claim 1 or 18. 5 Appeal2018-001499 Application 14/025,722 DECISION We reverse the Examiner's rejection of claims 1-34. REVERSED 6 Copy with citationCopy as parenthetical citation