Ex Parte Kanaujia et alDownload PDFPatent Trial and Appeal BoardDec 20, 201713843455 (P.T.A.B. Dec. 20, 2017) Copy Citation 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. 13/843,455 03/15/2013 Atul Kanaujia OV-lll 9926 74712 7590 12/21 Muir Patent Law, PLLC P.O. Box 1213 9913 Georgetown Pike, Suite 200 Great Falls, VA 22066 EXAMINER NGUYEN, KATHLEEN V ART UNIT PAPER NUMBER 2486 NOTIFICATION DATE DELIVERY MODE 12/22/2017 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): eofficeaction @ appcoll.com pto@muirpatentlaw.com PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE PATENT TRIAL AND APPEAL BOARD Ex parte ATUL KANAUJIA, NIELS HAERING, and MUN WAI LEE Appeal 2017-007880 Application 13/843,455 Technology Center 2400 Before ELENI MANTIS MERCADER, NORMAN H. BEAMER, and ADAM J. PYONIN, Administrative Patent Judges. MANTIS MERCADER, Administrative Patent Judge. DECISION ON APPEAL STATEMENT OF THE CASE Appellants appeal under 35 U.S.C. § 134(a) from the Examiner’s Final Rejection of claims 1—32, 35—37, and 43—50. Claims 33, 34 and 38-42 have been cancelled. We have jurisdiction under 35 U.S.C. § 6(b). We affirm-in-part. THE INVENTION Appellants’ claimed invention is directed to “model[ling] the shape of a human object monitored by a video stream” in which “3D human models, such as a coarse 3D human model and a detailed 3D human model may be Appeal 2017-007880 Application 13/843,455 estimated by mapping individual body part components to a frame” (Abstract). Independent claims 1, 32, and 45, reproduced below, are representative of the subject matter on appeal: 1. A method of video analysis, comprising: receiving plural video streams, each video steam providing a series of video images of a monitored location; detecting a human object within the video images; determining a three dimensional (3D) hull corresponding to the human object within the video images; determining initial pose hypotheses of the human object within the video images, the initial pose hypotheses comprising a plurality of pose predictions associated with a particular instant in time, each of the plurality of pose predictions associated with a corresponding first probability value; mapping a plurality of simple geometric volumes to each of the pose predictions of the initial pose hypotheses to obtain a plurality of corresponding coarse 3D human models; comparing each of the corresponding coarse 3D human models to the 3D hull; in response to the comparing of each of the corresponding coarse 3D human models to the 3D hull, modifying pose predictions of the initial pose hypotheses to obtain refined pose hypotheses; for each of the pose predictions of the refined pose hypotheses, mapping at least one standard human model 2 Appeal 2017-007880 Application 13/843,455 to a coarse 3D human model associated with the pose prediction to obtain an associated refined 3D human model; comparing each of the refined 3D human models to at least one of the video images to select an optimum 3D human model as representing the human object detected within the video images; and based on the selected optimum 3D human model, automatically detecting at least one of an event represented within at least one of the video images, a characteristic of the human object and a second object within at least one of the video images. 32. A method of video analysis, comprising: receiving plural video streams, each video steam providing a series of video images of a monitored location; detecting a human object within the video images; determining a coarse 3D human model representing the human object; and mapping at least one standard human model to the coarse 3D human model on a human part by part basis to obtain a refined 3D human model comprising mapping a body part of the at least one standard human model to a corresponding portion of the coarse 3D human model, the mapping of the body part including individual and separate adjustment of a size of the body part of the at least one standard human model, based on the mapping, automatically detecting at least one of an event represented within at least one of the video images, a characteristic of the human object and a second object within at least one of the video images. 3 Appeal 2017-007880 Application 13/843,455 45. A method of video analysis, comprising: receiving plural video streams, each video stream providing a series of video images of a monitored location; detecting a human object within the video images; determining a pose of the human object within the video images; mapping a plurality of simple geometric volumes to the pose to obtain a coarse 3D human-accessory combined model; mapping at least one standard human model to the coarse 3D human-accessory combined model associated with the pose to obtain an associated refined 3D human- accessory combined model; and based on the mapping of the at least one standard human model to the coarse 3D human-accessory combined model, automatically detecting at least one of an event represented within at least one of the video images, a characteristic of the human object and a second object within at least one of the video images, wherein the coarse 3D human-accessory combined model comprises a model of a human and a portable container or portable device. REFERENCES The prior art relied upon by the Examiner in rejecting the claims on appeal is the following: Monachino US 2005/0002561 A1 Jan. 6, 2005 4 Appeal 2017-007880 Application 13/843,455 Sundaresan US 2009/0232353 A1 Sept. 17, 2009 Black US 2010/0111370 A1 May 6, 2010 REJECTIONS The Examiner made the following rejections: Claims 1—26, 32, 35—37, 43, and 44 stand rejected under 35 U.S.C. § 103(a) as being unpatentable over Sundaresan in view of Black. Final Act. 4. Claims 27—31 and 45—50 stand rejected under 35 U.S.C. § 103(a) as being unpatentable over Sundaresan, in view of Black and Monachino. Final Act. 33. ISSUES The issues are whether the Examiner erred in finding that: 1. the combination of Sundaresan and Black teaches or suggests the limitation of in response to the comparing of each of the corresponding coarse 3D human models to the 3D hull, modifying pose predictions of the initial pose hypotheses to obtain refined pose hypotheses, as recited in claim 1; 2. the combination of Sundaresan and Black teaches or suggests the limitation of mapping at least one standard human model to the coarse 3D human model on a human part by part basis to obtain a refined 3D human model comprising mapping a body part of the at least one standard human model to a corresponding portion of the coarse 3D human model, the mapping of the body part 5 Appeal 2017-007880 Application 13/843,455 including individual and separate adjustment of a size of the body part of the at least one standard human model, as recited in claim 32; and 3. the combination of Sundaresan, Black, and Monachino teaches or suggests the limitation of “mapping a plurality of simple geometric volumes to the pose to obtain a coarse 3D human-accessory combined model,” as recited in claim 45. ANALYSIS Claim 1 Appellants argue Examiner error because the Examiner’s Answer “now appears to have changed the mapping of claim 1 to map the claimed plurality of pose predictions to the different ‘segment matches’ discussed at paragraph [0162] of Sundaresan” (Reply Br. 3—4), but under this mapping “Sundaresan fails to teach ‘mapping a plurality of simple geometric volumes to each of the pose predictions [segment matches] ’ as would be further required by claim 1” (Reply Br. 5). Appellants further contend “the Examiner’s Answer does not even attempt to explain how Sundaresan would meet these additional requirements of claim 1 when mapping Sundaresan’s segment matches to claim l’s pose predictions” (Reply Br. 5). We are persuaded by Appellants. In the Answer, the Examiner finds the claimed “determining initial pose hypotheses” is taught by Sundaresan because Sundaresan “discloses segmented body parts are being connected to generate a skeleton model” (Ans. 3, citing Sundaresan 1162), and that “[w]hen different segments are matched to create a skeleton model, each created skeleton model will have [a] corresponding probability of 6 Appeal 2017-007880 Application 13/843,455 registration” (Ans. 3, citing Sundaresan 1162, and equations 19 and 20). The Examiner additionally finds Sundaresan “discloses that stature is estimated from the initial skeleton model” (Ans. 3, citing Sandaresan 121—122). The results of the segmented body calculations are input to the procedure of Figure 10 at block 100. See Sandaresan 1122. In the Final Action, the Examiner finds the claimed “mapping a plurality of simple geometric volumes to each of the nose predictions of the initial hypotheses to obtain a plurality of corresponding coarse 3D human models” is taught by Sundaresan in Figure 10, in which “a super-quadric model is estimated in block 106 using the skeleton model” (Final Act. 5, citing Sundaresan Fig. 10,1124; see also Figs. 4, 3F, 5A—C, 109-110, emphasis added). Under any acceptance of the Examiner’s findings, in the claimed “determining initial pose hypotheses” step, “the initial pose hypotheses comprising a plurality of pose predictions” are poses that are changed in Sundaresan block 104, before being input to the super-quadric model in block 106. See Sundaresan Fig. 10, 122—123. Also, under the Examiner’s findings, the claimed “modifying pose predictions of the initial hypotheses” occurs in Sundaresan block 108. But block 108 is not “modifying pose predictions of the initial hypotheses” — block 108 is modifying poses that are determined in block 104 that are not the initial pose predictions. Neither Black nor Monachino are cited by the Examiner for curing this deficiency. Thus, we do not find the combination of Sundaresan and Black teaches or suggests “in response to the comparing of each of the corresponding coarse 3D human models to the 3D hull, modifying pose 7 Appeal 2017-007880 Application 13/843,455 predictions of the initial pose hypotheses to obtain refined pose hypotheses.” We are constrained by the record to reverse the Examiner’s rejection of independent claim 1, as well as claims 2—31 dependent on claim 1.1 Claim 32 Appellants argue Examiner error because “the Examiner now takes a [] position that Sundaresan does in fact teach individual and separate adjustment of a size of the body part of at least one standard human model, highlighting paragraph [0175] of Sundaresan” (Reply Br. 7, emphasis in original). Appellants contend that “Sundaresan is consistent with the disclosure of the present application and refers 3D volumetric models as a ‘human body model’ but does not consider its skeleton model as a human body model” (Reply Br. 9, quoting Sundaresan 1164). We are not persuaded by Appellants. The Examiner finds, and we agree, that “[individual changes in model parameters [of the model and pose parameters disclosed in Sundaresan 1175] will result in individual changes in corresponding tapered super-quadrics being mapped” (Ans. 5). The Examiner also finds, and we agree, that “Black is used to teach [a] standard human model by matching a fitted body model to a database 208 that contain stored 3D body models using a body matching component as in Paragraph [0066]” (Ans. 5) and further teaches a “body model [] segmented into body parts” (Ans. 5, citing Black || 205—206). 1 Should there be further prosecution, the Examiner should consider whether Sundaresan’s model used as a “starting point” to estimate the quadric model has fitting errors equivalent to association with a corresponding first probability values. 8 Appeal 2017-007880 Application 13/843,455 Appellants attack Sundaresan individually and do not persuade us the combined teachings of Sundaresan and Black fail to teach or suggest the claim limitation, under the Examiner’s finding in which Sundaresan’s method of processing data wherein changes in model parameters results in changes in tapered super-quadrics, is combined with the standard and segmented body models of Black. One cannot show nonobviousness by attacking references individually where, as here, the rejections are based on combinations of references. In re Keller, 642 F.2d 413, 426 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 1097 (Fed. Cir. 1986). Accordingly, we sustain the Examiner’s rejection of independent claim 32, and sustain the Examiner’s rejection of dependent claims 35 and 36 for similar reasons (see Reply Br. 10—11). Claim 45 Appellants argue Examiner error because Monachino “teaches detecting and dividing two dimensional blocks of grayscale values” using “the size, location and aspect ratio of such block,” but “[tjhese identified characteristics of size, location and aspect ratio do not constitute any model of anything in the analyzed image of Monachino” (Reply Br. 12, citing Monachino 141). Appellants further contend Monachino fails to teach a “combined human-accessory model” and fails to suggest modification of Sundaresan and Black to achieve the claimed invention (Reply Br. 12) because Monachino is “directed to dividing a portion of the image representing an object of interest to save it as a separate file” (Reply Br. 12). We are not persuaded of error. The Examiner finds, and we agree, that Monachino teaches “identifying candidate objects including carried 9 Appeal 2017-007880 Application 13/843,455 objects such as bags or suitcases, that the candidate objects are identified by one or more pattern recognition systems which can map a pattern of accessory object volume into the object to recognize the object” (Ans. 6, citing Monachino 18—19, which describe examining frames to identify objects such as bags or suitcases). The Examiner also finds, and we agree, that the incorporation of Monachino’s teachings into the teachings of Sundaresan and Black would “incorporate mapping an accessory simple geometric volume [Monachino] to a pose of the corresponding pose prediction [Sundaresan and Black] to have a 3D human with accessory model, automatically detecting a characteristic of a second object” with motivation supplied by Monachino (Final Act. 42, citing Monachino 17; see also Ans. 6—7, citing Monachino ^fl[ 24—25). Appellants’ arguments focus on the image processing teachings of Monachino and do not challenge the Examiner’s findings that Monachino teaches identifying carried objects such as bags or suitcases, along with identifying body areas, and the desirability of incorporating those teachings into the combined teachings of Sundaresan and Black for a combined human-accessory model. Accordingly, we sustain the Examiner’s rejection of independent claim 45. Appellants make additional arguments regarding dependent claims 46-48 (see Reply Br. 12—13) that purport to identify assertions by the Examiner regarding individual references, and these arguments are similarly unpersuasive as the arguments do not consider the combination of the references. Accordingly, we sustain the Examiner’s rejection of claims 46— 48, and claims 49-50 not separately argued (see App. Br. 16—22). 10 Appeal 2017-007880 Application 13/843,455 CONCLUSION The Examiner: 1. erred in finding the combination of Sundaresan and Black teaches or suggests the limitation of in response to the comparing of each of the corresponding coarse 3D human models to the 3D hull, modifying pose predictions of the initial pose hypotheses to obtain refined pose hypotheses, as recited in claim 1; 2. did not err in finding the combination of Sundaresan and Black teaches or suggests the limitation of mapping at least one standard human model to the coarse 3D human model on a human part by part basis to obtain a refined 3D human model comprising mapping a body part of the at least one standard human model to a corresponding portion of the coarse 3D human model, the mapping of the body part including individual and separate adjustment of a size of the body part of the at least one standard human model, as recited in claim 32; and 3. did not err in finding the combination of Sundaresan, Black, and Monachino teaches or suggests the limitation of “mapping a plurality of simple geometric volumes to the pose to obtain a coarse 3D human- accessory combined model,” as recited in claim 45. DECISION The Examiner’s decision rejecting claims 1—31 is reversed. The Examiner’s decision rejecting claims 32, 35—37, and 43—50 is affirmed. 11 Appeal 2017-007880 Application 13/843,455 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-IN-PART 12 Copy with citationCopy as parenthetical citation