`
`
`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`____________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`____________________
`
`SAMSUNG ELECTRONICS CO., LTD.; AND
`SAMSUNG ELECTRONICS AMERICA, INC.
`Petitioner
`
`v.
`
`IMAGE PROCESSING TECHNOLOGIES, LLC
`Patent Owner
`
`____________________
`
`IPR2017-01190
`Patent No. 6,717,518
`____________________
`
`PETITIONER’S REPLY
`
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`
`
`TABLE OF CONTENTS
`
`I.
`II.
`
`B.
`
`C.
`
`Contents
`INTRODUCTION ........................................................................................... 1
`ARGUMENT ................................................................................................... 2
`A.
`Claim Construction ............................................................................... 2
`1.
`“Histogram” ................................................................................ 2
`2.
`“anthropomorphic model” .......................................................... 3
`3.
`“characteristic of the face” / “facial characteristic” .................... 3
`4.
`“Selecting pixels of the portion of the image having
`characteristics corresponding to the feature to be
`detected”; “forming at least one histogram of the selected
`pixels” ......................................................................................... 5
`Ground A: Eriksson and Stringa Render Claim 39 Obvious ............. 12
`1.
`Eriksson and Stringa both disclose a “histogram” .................... 12
`2.
`Eriksson and Stringa both disclose forming a histogram
`“of the selected pixels” ............................................................. 18
`A POSA Would Have Been Motivated to Combine
`Stringa and Eriksson ................................................................. 23
`Ground B: Ando and Suenaga Render Claim 39 Obvious ................ 25
`1.
`Ando and Suenaga disclose forming a histogram of the
`selected pixels ........................................................................... 25
`Suenaga discloses a facial characteristic other than the
`feature to be detected ................................................................ 27
`A POSA Would Have Been Motivated to Combine Ando
`and Suenaga .............................................................................. 28
`III. CONCLUSION .............................................................................................. 30
`
`3.
`
`2.
`
`3.
`
`i
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`LIST OF EXHIBITS1
`Description
`
`Ex.
`No.
`1001 U.S. Patent No. 6,717,518 (“the ’518 Patent”)
`
`1002 Declaration of Dr. John C. Hart
`
`1003 Curriculum Vitae for Dr. John C. Hart
`
`Filing/Service
`Status
`Filed and served
`on 03/29/2017
`Filed and served
`on 03/29/2017
`Filed and served
`on 03/29/2017
`1004 Prosecution File History of U.S. Patent No. 6,717,518 Filed and served
`on 03/29/2017
`Filed and served
`on 03/29/2017
`
`1005 Martin Eriksson et al., Eye Tracking For Detection Of
`Driver Fatigue, IEEE Conference on Intelligent
`Transportation Systems (Nov. 1997) (“Eriksson”)
`1006 Luigi Stringa, Eyes Detection For Face Recognition,
`Applied Artificial Intelligence (1993) (“Stringa”)
`1007 U.S. Patent No. 5,805,720, Facial Image Processing
`System (Filed Mar. 11, 1996) (“Suenaga”)
`1008 U.S. Patent No. 5,293,427, Eye Position Detecting
`System and Method Therefor (Filed Dec. 11, 1991)
`(“Ueno”)
`1009 U.S. Patent No. 5,008,946, System For Recognizing
`Image (Filed Sept. 9, 1988) (“Ando”)
`1010 Declaration of William Garrity from U.C. Davis
`Regarding Stringa
`1011 Declaration of Dr. Umit Ozguner Regarding Eriksson
`
`1012 Excerpts from the Infringement Expert Report of Dr.
`Alan C.
`Bovik
`1013 [Proposed] Protective Order
`
`Filed and served
`on 03/29/2017
`Filed and served
`on 03/29/2017
`Filed and served
`on 03/29/2017
`
`Filed and served
`on 03/29/2017
`Filed and served
`on 03/29/2017
`Filed and served
`on 03/29/2017
`Filed and served
`on 08/03/2017
`
`Filed and served
`on 08/03/2017
`
`
`
` 1
`
` Citations to non-patent publications are to the original page numbers of the
`
`publication, and citations to U.S. patents are to column:line number of the patents.
`
`ii
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`1014 Redline Comparison of [Proposed] Protective Order
`
`1015 Deposition of Gerard P. Grenier (Supplemental
`Evidence regarding Exhibit 1005)
`1016 Declaration of Gerard P. Grenier (Supplemental
`Evidence regarding Exhibit 1005)
`1017 Martin Eriksson et al., Eye Tracking For
`Detection Of Driver Fatigue - Abstract
`(Supplemental Evidence regarding Exhibit 1005)
`1018 Martin Eriksson et al., Eye Tracking For
`Detection Of Driver Fatigue - Abstract
`(Supplemental Evidence regarding Exhibit 1005)
`1019 Martin Eriksson et al., Eye Tracking For
`Detection Of Driver Fatigue, IEEE Conference
`on Intelligent Transportation Systems (Nov.
`1997) (Supplemental Evidence regarding Exhibit
`1005)
`
`Filed and served
`on 08/03/2017
`Served on
`11/01/2017
`Served on
`11/01/2017
`Served on
`11/01/2017
`
`Served on
`11/01/2017
`
`Served on
`11/01/2017
`
`iii
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`I.
`
`INTRODUCTION
`The Board instituted review of the ’518 Patent on two grounds: A) Claim 39
`
`is obvious over Eriksson and Stringa; and B) Claim 39 is obvious over Ando and
`
`Suenaga. Paper 11 at 26.
`
`Regarding Ground A, Patent Owner Image Processing Technologies (“IPT”)
`
`relies on attorney argument alone to assert that Erikson and Stringa do not disclose
`
`“histograms.” But IPT has already repeatedly admitted in this proceeding that the
`
`data representations cited in these references are histograms. Further, Eriksson, a
`
`peer reviewed, IEEE publication, expressly calls the cited data representations
`
`“histograms.”
`
`For Ground B, IPT argues that Suenaga does not disclose a “facial
`
`characteristic” other than the feature to be detected, but does not dispute that the
`
`primary Ground B reference, Ando, discloses this limitation.
`
`IPT is thus left to argue that none of the references disclose Claim 39,
`
`because they all select and form histograms of all pixels in a particular area, rather
`
`than only pixels of the feature of the eye (iris, pupil, or cornea) being detected.
`
`This argument relies on an interpretation of the claim that has already been rejected
`
`by the Board and is contrary to the specification. Regardless, IPT’s argument
`
`ignores disclosures in the references that plainly satisfy the claim, even under
`
`IPT’s rejected interpretation.
`
`1
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`II. ARGUMENT
`A. Claim Construction
`1.
`“Histogram”
`The Board need not construe “histogram” because both instituted grounds
`
`include at least one reference that expressly discloses the use of a “histogram,” as
`
`required by Claim 39. For example, for Ground A, Eriksson labels the relevant
`
`“histograms” as such:
`
`
`
`Ex. 1005 at 318 (text in original, emphasis added). Given the express use of the
`
`term “histogram” by this IEEE peer reviewed publication, there can be no dispute
`
`that Eriksson discloses the required “histogram,” particularly given the broadest
`
`reasonable interpretation (“BRI”) standard applicable in this review.2 For Ground
`
`
`
` 2
`
` The second Ground A reference, Stringa, also discloses the claimed use of a
`
`histogram but uses the term “grey-level distribution.” See Section II.B.1.
`
`2
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`B, IPT does not dispute the disclosure of a “histogram” by either reference.
`
`Even if the Board were to grant IPT’s request (Paper 15 at 28–29) and
`
`accept the District Court’s construction for “histogram” (rendered under Philips
`
`rather than the BRI), IPT’s application of that construction is flawed. See Section
`
`II.B.1.
`
`2.
`“anthropomorphic model”
`The Board need not interpret “anthropomorphic model” since IPT does not
`
`advance any arguments based on this term.
`
`3.
`“characteristic of the face” / “facial characteristic”
`IPT argues that construing these terms will assist in determining whether
`
`Eriksson (Ground A) and Suenaga (Ground B) invalidate Claim 39. Paper 15 at
`
`32. Notwithstanding its stated purpose, IPT makes no arguments related to
`
`Eriksson. See Paper 15 at 44–57. Thus, this element is undisputed for Ground A.
`
`For Ground B, IPT argues only that Suenaga’s centroid of the face cannot
`
`qualify as a “facial characteristic.” Id. at 66. However, IPT does not dispute that
`
`Ando, the primary Ground B reference, discloses this element. Id. Thus, there is
`
`no need for the Board to consider IPT’s construction or related arguments.
`
`Regardless, IPT’s proposed construction is vague and unnecessary. The
`
`plain language of the claim is sufficiently clear—“characteristic” and “face” are
`
`not terms of art—and the phrase “distinguishing element” in IPT’s construction is
`
`3
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`no more clear than the claim term “characteristic.”
`
`Further, applying the BRI, IPT’s construction, to the extent it is understood,
`
`appears overly narrow. The specification uses the term “characteristic” more
`
`broadly than IPT suggests, including to refer to parameters such as velocity,
`
`direction, luminance, hue and saturation:
`
`While the invention is being described with respect to identification of
`the nostrils as a starting point to locating the eyes, it is foreseen that any
`other facial characteristic, e.g., the nose, ears, eyebrows, mouth, etc.,
`and combinations thereof, may be detected as a starting point for
`locating the eyes. These characteristics may be discerned from any
`characteristics capable of being searched by the system, including
`CO, DP, velocity, direction, luminance, hue and saturation.
`Ex. 1001 at 29:52–60 (emphasis added). The specification similarly indicates that
`
`a “facial characteristic” allows use of an anthropomorphic model to find the feature
`
`to be detected:
`
`In order to select the sub-area of the image, the controller interacts with
`the histogram formation unit to identify the head of the driver in the
`image, or a facial characteristic of the driver, such as the driver's
`nostrils. The controller then identifies the sub-area of the image using
`an anthropomorphic model.
`Id. at 3:54–59 (emphasis added).
`
`Thus, under the BRI, if these terms are construed, they should not be limited,
`
`e.g., to the nose, but should include other aspects of the face that allow use of an
`
`4
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`anthropomorphic model to find the feature to be detected.
`
`Lastly, IPT argues that the patent discloses “alternative” embodiments based
`
`on: (i) “head frame” and (ii) “facial characteristics.” Paper 15 at 16-27. However,
`
`IPT took a directly contrary position in litigation in its infringement expert report,
`
`arguing that the “facial characteristic” requirement of Claim 39 is satisfied by the
`
`location of the head (the “bounding box”). Ex. 2012 at 51, 55.
`
`
`
`Id. IPT does not even attempt to explain how a “bounding box” is different from a
`
`“head frame” or a centroid.
`
`4.
`
`“Selecting pixels of the portion of the image having
`characteristics corresponding to the feature to be detected”;
`“forming at least one histogram of the selected pixels”
`IPT repeats the argument made in its Preliminary Response (which the
`
`5
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`Board rejected)—that 39[d] “‘selecting pixels of a portion of the image having
`
`characteristics corresponding to the feature to be selected’ precludes selection of
`
`pixels that are not of the feature itself.” Paper 11 at 16. IPT now expands on this
`
`rejected interpretation to further assert that 39[e] “forming at least one histogram
`
`of the selected pixels” cannot include “forming a histogram of all pixels in…a
`
`rectangular area” but must include “only iris, pupil, or cornea pixels.” See, e.g.,
`
`Paper 15 at 34, 57.
`
`A linchpin of IPT’s argument is that 39[c] (“identifying a portion of the
`
`image”) uses the term “comprising,” and 39[d] (“selecting pixels”) and 39[e]
`
`(“forming at least one histogram”) do not. Id. at 37. Based on that difference, IPT
`
`implies that the terms “having” in 39[d] and “of” in 39[e] must be closed
`
`transitional terms. Id. But IPT’s argument ignores the use of “comprising” as the
`
`transitional term in the preamble, which “signals that the entire claim is
`
`presumptively open-ended.” Manual of Patent Examining Procedure §2111.03
`
`(9th ed. Rev. 8, 2017) (quoting In Gillette Co. v. Energizer Holdings Inc., 405 F.3d
`
`1367, 1371–73, 74 USPQ2d 1586, 1589–91 (Fed. Cir. 2005)) (emphasis added).
`
`Applying this well-settled claim construction principle and the BRI, it is plain that
`
`39[d] and [e] are open-ended. To hold otherwise would improperly rewrite the
`
`claim as follows:
`
`6
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`[d] selecting only those pixels of the portion of the image having
`characteristics corresponding to the feature to be detected;
`[e] forming at least one histogram of only the selected pixels;
`
`
`
`Even if IPT were correct in ignoring the preamble, its analysis of 39[d] and
`
`[e] is flawed:
`
`“Non-conventional transitional phrases (i.e., other than “comprising,”
`“consisting essentially of,” and “consisting”) are interpreted in light of
`the specification to determine whether open or closed claim language
`is intended. See, e.g., Lampi Corp. v. American Power Prods. Inc., 228
`F.3d 1365, 1376 (Fed. Cir. 2000) (interpreting “having” as open
`terminology, allowing the inclusion of other components in addition
`to those recited) …
`Ex parte Jeffrey Santrock, 2017 Pat. App. LEXIS 4143, *9–10 (P.T.A.B. Apr. 28,
`
`2017) (emphasis added). In Santrock, the Board found the transitional term to be
`
`open because it found “nothing in the Specification to suggest that” the transitional
`
`term “was intended as other than open claim language, allowing but not requiring
`
`additional, unrecited elements.” Id. Similarly, here, nothing in the specification
`
`(i) requires the “selecting” 39[d] and “forming” 39[e] to include only pixels of the
`
`iris, pupil, or cornea, or (ii) precludes these limitations from selecting and forming
`
`a histogram of all pixels in a specific area (“portion”) identified in 39[c].
`
`For example, IPT agrees that the “histogram” is not limited to intensity
`
`values, but may be of any parameter such as speed, velocity, movement, or
`
`7
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`shadows. Paper 15 at 30, 24. But none of these parameters are unique to features
`
`of the eye and all will capture other facial features. As shown, for example, in
`
`Figures 27, 32, 33, 34, and 36, the histograms of the patent include pixels other
`
`than those for the feature of interest.
`
`
`
`
`
`
`
`8
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`Petitioner’s Reply in IPR2017-01190 (US. Patent No. 6,717,518)
`
`center
`Xn'Yn
`
`
`
`
`
`
`
`
`
`9
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`
`
`Ex. 1001 at Figs. 36, 27, 32–34 (red arrows added). As an example, Figure 36
`
`shows two histograms of a pupil, and the histograms both include pixels that are
`
`not pupil pixels (annotated in red). Likewise, the specification includes explicit
`
`examples where all pixels in an area are selected for forming a histogram. Id. at
`
`18:58–19:25.
`
`Lastly, IPT’s arguments are contrary to positions it has taken in litigation.
`
`As shown in IPT’s infringement expert report, IPT argued that 39[c] (“identifying
`
`a portion of the image of the face comprising the feature to be detected…”) is
`
`satisfied by identifying an “eye sub-region” within a detected face:
`
`10
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`
`
`Ex. 1012 at 56–58 (red outline in original). IPT also argued that the selecting and
`
`forming a histogram of the selected portion, as recited in 39[d] and [e], is satisfied
`
`by forming a histogram of all pixels within this same “eye sub-region”:
`
`
`
`
`
`Id. at 58–62 (red outline in original). Thus, contrary to its current construction, in
`
`litigation, IPT argued that the 39[d] and [e] selecting and forming limitations
`
`encompassed forming a histogram of all pixels within the 39[c] identified
`
`“portion” (indicated by the red box) and may include pixels that are not the iris,
`11
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`pupil, or cornea (see, e.g., left most pink boxes in above figure).
`
`Applying the BRI as required here, IPT’s construction should be rejected
`
`and the plain and ordinary meaning adopted.
`
`B. Ground A: Eriksson and Stringa Render Claim 39 Obvious
`1.
`Eriksson and Stringa both disclose a “histogram”
`Based on attorney argument alone, IPT maintains that Figure 5 of Eriksson
`
`does not show a histogram, but instead is a “plot of intensity values.” Paper 15 at
`
`47. But Eriksson, a peer reviewed, IEEE publication, specifically describes Figure
`
`5 as showing “histograms”:
`
`
`
`Ex. 1005 at 318 (text in original, emphasis added). Thus, Figure 5 has already
`
`been confirmed by a POSA to illustrate histograms. In fact, IPT repeatedly
`
`admitted in its Preliminary Response that Figure 5 discloses a “histogram”:
`
`Eriksson discloses forming and analyzing a histogram of all pixels in
`an area…Eriksson forms an intensity histogram of all the pixels…The
`region for which the histogram is formed is the rectangular area around
`the estimated position of the iris…This histogram is analyzed
`12
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`over time…Figure 5 of Eriksson shows this histogram.
`
`Paper 6 at 30 (emphases added).
`
`IPT similarly relies on attorney argument alone to argue Stringa does not
`
`disclose “histograms” but instead discloses “horizontal grey-level distributions.”
`
`Paper 15 at 48. IPT is again incorrect. Stringa uses the phrase “distribution”
`
`generically, including when referring to “horizontal grey-level histograms” as
`
`“horizontal grey-level distributions.” For example, the text describing Figure 3
`
`uses the term “histogram” while the caption to Figure 3 uses the term
`
`“distribution” to refer to the same data representation. Paper 15 at 370. Indeed,
`
`IPT repeatedly admitted that Stringa’s “horizontal grey-level distribution” is a
`
`“histogram” in its Preliminary Response:
`
`Stringa forms a horizontal grey-level histogram…Stringa uses the
`second derivative of
`this histogram
`to detect an eye pupil
`location…Stringa, like Eriksson, requires that all pixels be used in the
`histogram…Stringa relies on the shape of the second derivative curve
`of the histogram that is based on grey-level values of all pixels in the
`zone to identify the pupil location.
`Paper 6 at 31–32 (emphases added).
`
`Review of the references themselves also plainly contradicts IPT’s
`
`unsupported current argument that these references disclose only “plots of the
`
`intensity values” for pixels on a particular line rather than “histograms.” Paper 15
`
`at 47-48. Stringa includes a “Preliminaries” section that “introduce(s) some
`13
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`general notions that will be used throughout.” Ex. 1006 at 370. One of the
`
`“general notions” is that of vertical and horizontal histograms, labeled “H(y)” and
`
`“H(x)” in Figure 3. The Preliminaries section explains how these histograms are
`
`formed using a leading edge or grey level thresholding formula for each pixel. Id.
`
`at 371. In a leading edge example, each pixel is assigned a binary value depending
`
`on a difference in grey value between adjacent pixels:
`
`Id. To create a vertical histogram H(y), for each point y on the vertical axis,
`
`Stringa takes the sum of all leading edges (“1” pixels) of the horizontal line at that
`
`
`
`y location:
`
`Id. at 372. The resulting leading edge vertical histogram H(y) is shown in Figure
`
`
`
`3:
`
`14
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`
`
`Id. at 370 (annotation added).
`
`Although explained in the context of a leading edge vertical histogram H(y),
`
`this explanation applies equally to a leading edge horizontal histogram, such as
`
`H(x) shown in Figure 3, and a grey-level vertical or horizontal histogram (or
`
`“grey-level distribution”). Ex. 1006 at 370, Figure 3 (referring to “filtered edges
`
`and grey level distributions”). In all cases, the histogram meets IPT’s proposed
`
`construction of the term: “a statistical representation of the frequency of
`
`occurrence with which values of a parameter fall within a series of intervals.”
`
`Paper 15 at 28. The parameter would be the thresholded grey-level or the
`
`thresholded leading edge, and the series of intervals would be the x or y location.
`15
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`Turning to the specific disclosures relevant to claim 39, Stringa forms a
`
`horizontal grey level distribution (histogram) of pixels along the x-axis in the
`
`expectation zone. Ex. 1006 at 377–78. At each position along the x-axis, the
`
`histogram counts the number of pixels in the corresponding vertical line that
`
`exceed a certain intensity threshold. See, e.g., Ex. 1006 at 371 (describing an
`
`example thresholding formula for calculating a histogram); Ex. 1005 at ¶105. This
`
`“horizontal grey-level distribution” is the “histogram” required by Claim 39, as
`
`IPT admits. Paper 6 at 31–32.
`
`IPT ignores its prior admissions and now argues that Stringa’s “horizontal
`
`grey-level distribution” is not a histogram, but “is a smoothed value of intensity for
`
`pixels in one horizontal line of the image area.” Paper 15 at 48. By “smoothed,”
`
`IPT seems to be referring to the “band-pass” filtering performed to create the
`
`function GYrs(x). Id. However, GYrs(x) is not the histogram—it is calculated based
`
`on “an analysis of” the histogram. Ex. 1006 at 377. It is thus clear that IPT’s
`
`original admission, that the “horizontal grey-level distribution” is a “histogram,” is
`
`correct.
`
`Eriksson builds on the teachings of Stringa (Ex. 1005 at 315, “One
`
`interesting application for face recognition was developed by Stringa [12]…We
`
`use a similar approach…”) and uses this same terminology (“horizontal
`
`histogram”) to describe the same data construct (Figure 5), in a section titled
`
`16
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`“Horizontal histogram across the pupil.” Ex. 1005 at 317–318. It thus discloses
`
`the claimed “histogram” for the same reason as Stringa.
`
`Even if IPT’s new argument, that Eriksson and Stringa disclose “plots of the
`
`intensity values for pixels on a particular horizontal line” (Paper 15 at 47, 48), was
`
`correct, Dr. Hart explains why these references still disclose “histograms” under
`
`IPT’s construction. In testimony elicited by IPT, Dr. Hart explained that what
`
`would be counted is “the frequency of photons or other radiometric energy,
`
`radiometric power specifically.” Ex. 2003 at 143:14–144:23. Thus, Eriksson’s
`
`histograms would represent the frequency with which units of radiometric energy
`
`(the “parameter”) fall within a series of intervals, where each interval is a location
`
`on the x-axis, thereby satisfying IPT’s construction. IPT does not substantively
`
`respond to this argument, and instead simply declares, without support, that “[t]his
`
`argument proves too much, as anything done with intensity or color values (which
`
`necessarily are based on number of photons) would qualify…” Paper 15 at 29–30.
`
`In essence, IPT argues that plots of intensity might often qualify as histograms, but
`
`that is not a reason why they are not histograms.
`
`Lastly, IPT argues that Eriksson does not use a histogram but uses “match
`
`values” applied to three pixels along a line through the pupil to determine whether
`
`the eye is open or closed. Paper 15 at 45–46; Ex. 1005 at 318. But Eriksson
`
`explains that it “need[s] a robust way to determine if the eyes are open or closed;
`
`17
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`so we developed a method that looks at the horizontal histogram across the
`
`pupil.” Ex. 1005 at 317. Further, Eriksson includes a sub-section titled
`
`“Horizontal histogram across the pupil” in which it states “We use the
`
`characteristic curve generated by plotting the image-intensities along the line going
`
`through the pupil from left to right, as shown in Figure 5.” Id. at 318 (emphasis
`
`added). And Figure 5 is captioned “Histograms corresponding to an open and a
`
`closed eye, respectively.” Id. (emphasis added).
`
`As described in Eriksson, the matching function IPT cites is not used in
`
`place of the histogram, rather it is specifically used to analyze the histogram:
`
`When the eye is open, the valley in the intensity-curve [histogram]
`corresponding to the pupil will be surrounded by two large peaks
`corresponding to the sclera. When the eye is closed, this curve is
`usually very flat in the center…. This will lead to a good match when
`the eye is open, and very likely to a bad match when the eye is closed.
`Ex. 1005 at 318. Additionally, Dr. Hart explained at deposition that the parameters
`
`of the matching function are set using the histogram of Figure 5. Ex. 2003 at
`
`139:15–141:4. Thus, IPT’s argument that the Figure 5 “histograms” are “merely
`
`illustrative” is incorrect.
`
`2.
`
`Eriksson and Stringa both disclose forming a histogram “of
`the selected pixels”
`IPT argues that Eriksson and Stringa “do[] not teach or suggest forming a
`
`histogram of the selected pixels having characteristics of the iris, pupil, or cornea”
`
`18
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`because the histograms in these references are “formed from all the pixels in a
`
`particular area.” Paper 15 at 48–49. IPT’s argument hinges on its improper
`
`interpretation, which has already been rejected. Id. at 16. If the Board maintains
`
`its position, there is no dispute that these limitations are satisfied.
`
`Even if IPT’s construction was accepted (and the selection of pixels must be
`
`less than all pixels in the “portion” identified in step [c]), IPT’s arguments address
`
`only certain aspects of the cited references, and ignore other aspects cited in the
`
`Petition. One example of how Eriksson satisfies the claim is summarized below:
`
`Claim 39
`
`Eriksson
`
`“A process of detecting a feature of an
`eye, the process comprising the steps
`of” (39[pre])
`“acquiring an image of the face of the
`person, the image comprising pixels
`corresponding to the feature to be
`detected” (39[a])
`“identifying a characteristic of the face
`other than the feature to be detected”
`(39[b])
`
`The “feature” is a pupil. Paper 2 at 29.
`
`The image is acquired by a video
`camera and consists of pixels. Paper 2
`at 29.
`
`Eriksson identifies the horizontal
`center of the face. Paper 2 at 31.
`
`19
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`“identifying a portion of the image of
`the face comprising the feature to be
`detected using an anthropomorphic
`model based on the location of the
`identified facial characteristic” (39[c])
`
`“The search-space is now limited to the
`area around [the symmetry] line, which
`reduces the probability of having
`distracting features in the background.”
`Paper 2 at 33. For clarity, the
`“portion” is approximately drawn by
`Petitioner in red on Figure 1 below:
`
`“selecting pixels of the portion of the
`image having characteristics
`corresponding to the feature to be
`detected” (39[d])
`
`
`
`Eriksson selects pixels based on at least
`one characteristic corresponding to the
`pupil: (i) the pixels correspond to
`“intensity valleys” in the vertical
`gradient histogram, (ii) the pixels
`“must be located fairly close to the
`center of the face,” and (iii) the pixels
`should be on the same row. Paper 2 at
`36. These characteristics are used to
`define the vertical and horizontal
`bounds of the selected pixels, indicated
`by the red boxes in Figure 5 below.
`
`20
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`“forming at least one histogram of the
`selected pixels” (39[e])
`
`Eriksson forms a histogram of the
`selected pixels (indicated in red).
`Paper 2 at 37.
`
`
`
`“analyzing the at least one histogram
`over time to identify characteristics of
`the feature to be detected” (39[f])
`
`said feature being the iris, pupil or
`cornea. (39[g])
`
`
`The histogram is analyzed over time to
`determine if the eye is open or closed
`(i.e., the pupil being present or absent).
`Paper 2 at 39.
`Eriksson identifies the pupil. Paper 2
`at 40.
`
`The pixels “selected” (39[d]) and “formed” in the histogram (39[e]) are less than
`
`all of the pixels in the “portion” identified in 39[c], thus satisfying IPT’s
`
`application of the claim. IPT does not even address this read of Eriksson.
`
`For Stringa, the horizontal grey-level histogram (“horizontal grey-level
`
`distribution”) is formed using a threshold to determine whether a pixel is included
`
`in the histogram or not. Ex. 1006 at 370-372; see Section II.B.1. This selection
`
`based on luminance or grey-level is precisely the method IPT admits is used by
`
`the ’518 Patent to perform the selection function: “pixels with ‘very low
`
`luminance levels and high gloss’ are selected.” Paper 15 at 40–41. Thus, Stringa
`
`and Eriksson satisfy the claim under IPT’s construction.
`
`21
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`Even if the references selected and formed a histogram of all pixels in the
`
`area around the eye, as IPT argues, this would still satisfy the claim under the plain
`
`and ordinary meaning of the terms. The area around the eye in Figure 5 of
`
`Eriksson and the “expectation zone” of Stringa define pixels “having
`
`characteristics corresponding to the feature to be detected,” as required by 39[d]—
`
`i.e., x/y expected location (the positional “characteristic”) of the feature. Paper 2 at
`
`37. These positional characteristics are no different from the positional
`
`characteristics used to select pixels in the ’518 Patent. See, e.g., Ex. 1001 at
`
`18:58–19:25.
`
`IPT suggests that 39[f] is not satisfied because Eriksson does not “find[]” the
`
`iris, pupil, or cornea using a histogram, but instead uses a matching function
`
`applied to three intensity values along a horizontal line. Paper 15 at 49. This is a
`
`restatement of IPT’s argument that Eriksson does not “use” the Figure 5
`
`“histogram” and is wrong for the same reasons. See Section II.B.1 at 17–18. IPT
`
`further argues that the histogram cannot be used because the curve is
`
`“unpredictable” when the eye is closed. Paper 15 at 50. But that unpredictable
`
`shape is used by Eriksson to provide a “bad match when the eye is closed,”
`
`whereas the valley in the histogram provides “a good match when the eye is open.”
`
`Ex. 1005 at 318.
`
`Contrary to IPT’s strained reading, Eriksson’s use of the histogram to
`
`22
`
`
`

`

`Petitioner’s Reply in IPR2017-01190 (U.S. Patent No. 6,717,518)
`
`identify the pupil is the same as Figure 36 of the patent, which IPT admits
`
`embodies the claim. Paper 15 at 24. In Figure 36, the pupil is indicated by a peak
`
`in the histogram. In Eriksson, this same method is used, but is inverted so the
`
`pupil is indicated by a valley in the histogram:
`
`
`
`
`
`Ex. 1001 at Fig. 36 (left); Ex. 1005 at Fig. 5 (right).
`
`3.
`
`A POSA Would Have Been Motivated to Combine Stringa
`and Eriksson
`Based on attorney argument alone, IPT argues that it would not have been
`
`obvious to combine Eriksson and Stringa because of “the broad ranges of choices”
`
`for face detection algorithms. Paper 15 at 52. Such arguments have been roundly
`
`rejected: “The fact that a reference discloses a multitude of effective combinations
`
`does not render any particular formulation less obvious.” Ex parte Santrock, 2017
`
`Pat. App. LEXIS 4143 at *13–14 (collecting cases) (citations and quotations
`
`omitted). Notably, IPT only identifies three algorithms (Paper 15 at 53), whereas
`
`the case it cites included an infinite number of possibilities and held th