`Filed: June 29, 2023
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`
`
`
`
`LG ELECTRONICS INC.,
`Petitioner,
`v.
`JAWBONE INNOVATIONS, LLC,
`Patent Owner.
`
`IPR2023-01134
`U.S. Patent No. 11,122,357
`
`
`PETITION FOR INTER PARTES REVIEW
`OF CLAIMS 1-20 OF U.S. PATENT NO. 11,122,357
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`TABLE OF CONTENTS
`
`Page No.
`
`-i-
`
`C.
`
`D.
`
`E.
`
`B.
`
`C.
`
`INTRODUCTION ..................................................................................................... 1
`I.
`BACKGROUND ............................................................................................. 3
`A. Griffiths and Jim Publish Their Seminal GSC Article in 1982. ........... 3
`B. Over the Next Two Decades, the GSC was Used in Microphone
`Arrays to Reduce Noise. ........................................................................ 6
`Published in 2001, Brandstein Illustrates How to Use a GSC
`with a Microphone Array to Reduce Noise. .......................................... 7
`Contemporaneously, Gannot Taught Adapting the GSC to
`Handle Arbitrary Transfer Functions. ................................................. 10
`Other Concepts in the ’357 Patent Were Well Known. ...................... 13
`1.
`Filtering and Summing in the Time Domain Were Well
`Known. ...................................................................................... 13
`Delaying Signals Based on Geometry to Adjust for
`Differences in Arrival Times Was Well Known. ..................... 14
`THE ’357 PATENT ....................................................................................... 16
`A.
`The ’357 Patent Discloses Nothing Innovative. .................................. 16
`1.
`The ’357 Patent Purports to Distinguish Itself from the
`Prior Art by Using a Virtual Microphone Designed to
`Capture Only Noise, Which Had Been Known for
`Decades. .................................................................................... 16
`The ’357 Patent Concedes It Relies on Known
`Techniques to Form Virtual Microphones from Physical
`Microphones. ............................................................................. 16
`The ’357 Patent Discloses Formulas for the Purportedly
`Innovative Set of Virtual Microphones, But These
`Formulas Rely on Near-Field Design. ...................................... 18
`The Claims Recite Generic Virtual Microphones and Generic
`Signal Processing. ............................................................................... 20
`The Claims Were Not Carefully Scrutinized During
`Prosecution. ......................................................................................... 21
`
`II.
`
`2.
`
`2.
`
`3.
`
`
`
`TABLE OF CONTENTS
`(Cont’d)
`
`Page No.
`
`III. STATEMENT OF RELIEF REQUESTED .................................................. 22
`A. Grounds ............................................................................................... 22
`B.
`The Earliest Priority Date the ’357 Patent Claims Is June 13,
`2007. .................................................................................................... 22
`The References Are Prior Art. ............................................................. 22
`C.
`The Asserted References Are Analogous Art. .................................... 23
`D.
`IV. LEVEL OF ORDINARY SKILL .................................................................. 23
`V.
`CLAIM CONSTRUCTION .......................................................................... 24
`VI. GROUNDS OF UNPATENTABILITY ........................................................ 24
`A. Ground 1: Brandstein and Gannot ....................................................... 24
`1.
`Claim 1 ...................................................................................... 25
`a.
`Preamble ......................................................................... 25
`b.
`First Virtual Microphone Comprising a
`Combination of Signals from First and Second
`Physical Microphones ..................................................... 26
`Second Virtual Microphone ............................................ 27
`Substantially Similar Responses to Noise and
`Substantially Dissimilar Responses to Speech ............... 28
`A Signal Processor Operative to Combine
`Microphone Signals by Filtering and Summing in
`the Time Domain ............................................................ 31
`Applying a Varying Linear Transfer Function ............... 33
`f.
`Generating an Output Signal with Reduced Noise ........ 36
`g.
`Claim 2 ...................................................................................... 37
`Claim 3 ...................................................................................... 38
`Claim 4 ...................................................................................... 39
`Claim 5 ...................................................................................... 41
`a.
`Claim 5 Encompasses Standard Near-Field Time-
`Alignment. ...................................................................... 42
`
`c.
`d.
`
`e.
`
`2.
`3.
`4.
`5.
`
`
`
`
`
`TABLE OF CONTENTS
`(Cont’d)
`
`Page No.
`
`b.
`
`Brandstein Discloses or Renders Obvious
`Standard Near-Field Time-Alignments for the
`GSC. ................................................................................ 43
`Claim 6 ...................................................................................... 49
`6.
`Claim 7 ...................................................................................... 52
`7.
`Claim 8 ...................................................................................... 52
`8.
`Claim 9 ...................................................................................... 52
`9.
`10. Claim 10 .................................................................................... 53
`11. Claim 11 .................................................................................... 54
`12. Claim 12 .................................................................................... 56
`13. Claim 13 .................................................................................... 57
`14. Claim 14 .................................................................................... 58
`15. Claim 15 .................................................................................... 60
`a.
`Preamble ......................................................................... 60
`b.
`First Virtual Microphone ................................................ 60
`c.
`Second Virtual Microphone ............................................ 60
`d.
`Substantially Similar Responses to Noise and
`Substantially Dissimilar Responses to Speech ............... 61
`Virtual Microphone Array with a Single Null ................ 61
`e.
`Signal Processor.............................................................. 62
`f.
`Applying a Varying Linear Transfer Function ............... 63
`g.
`Generating an Output Signal with Reduced Noise ......... 63
`h.
`16. Claim 16 .................................................................................... 64
`17. Claim 17 .................................................................................... 64
`18. Claim 18 .................................................................................... 65
`19. Claim 19 .................................................................................... 65
`20. Claim 20 .................................................................................... 66
`B. Ground 2: Brandstein, Gannot, and Griffiths-Jim ............................... 66
`C. Ground 3: Brandstein, Gannot, and McCowan ................................... 68
`
`
`
`
`
`TABLE OF CONTENTS
`(Cont’d)
`
`Page No.
`
`VII. SECONDARY CONSIDERATIONS OF NONOBVIOUSNESS ............... 71
`VIII. DISCRETIONARY DENIAL UNDER §314(A) IS NOT
`APPROPRIATE. ........................................................................................... 72
`A.
`Co-Pending Litigation (Fintiv) ............................................................ 72
`1.
`Factor 1: Potential Stay ............................................................. 72
`2.
`Factor 2: Proximity of Trial to FWD ........................................ 72
`3.
`Factor 3: Investment in Parallel Proceeding ............................. 73
`4.
`Factor 4: Overlapping Issues .................................................... 73
`5.
`Factor 5: The Parties ................................................................. 75
`6.
`Factor 6: Other Circumstances .................................................. 75
`Prior IPR Petitions (General Plastic) ................................................... 76
`B.
`IX. DISCRETIONARY DENIAL UNDER §325(D) IS NOT
`APPROPRIATE. ........................................................................................... 77
`X. MANDATORY NOTICES ........................................................................... 78
`A.
`Real Parties-In-Interest (37 C.F.R. §42.8(b)(1)) ................................. 78
`B.
`Related Matters (37 C.F.R. §42.8(b)(2)) ............................................. 78
`C.
`Lead and Backup Counsel (37 C.F.R. §42.8(b)(3)) ............................ 80
`D.
`Service Information (37 C.F.R. §42.8(b)(4)) ...................................... 80
`E.
`Payment of Fees (37 C.F.R. §42.103) ................................................. 81
`F.
`Grounds for Standing (37 C.F.R. §42.104(a)) .................................... 81
`CONCLUSION ........................................................................................................ 81
`
`
`
`
`
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`TABLE OF AUTHORITIES
`
`Cases:
`
`Page(s):
`
`Amazon.com, Inc. v. Jawbone Innovations, LLC,
`IPR2023-00251, Paper 1 (P.T.A.B. Nov. 21, 2022) ---------------------------- 3
`Amazon.com, Inc. v. Jawbone Innovations, LLC,
`IPR2023-00251, Paper 15 (P.T.A.B. June 1, 2023) --------------------------- 72
`Apple Inc. v. Fintiv, Inc.,
`IPR2020-00019, Paper 11 (P.T.A.B. Mar. 20, 2020) --------------------- 72, 75
`Cal. Inst. of Tech. v. Broadcom Ltd.,
`25 F.4th 976 (Fed. Cir. 2022) ----------------------------------------------------- 74
`Central Security Group,
`IPR2019-01609, Paper 11 (P.T.A.B. Feb. 26, 2020) --------------------------- 76
`Celltrion, Inc. v. Genentech, Inc.,
`IPR2018-01019, Paper 11 (P.T.A.B. Oct. 30, 2018) -------------------------- 76
`In re GPAC Inc.,
`57 F.3d 1573 (Fed. Cir. 1995) ----------------------------------------------------- 23
`GAF Materials LLC v. Kirsch Research and Dev., LLC,
`IPR2021-00192, Paper 14 (P.T.A.B. May 25, 2021) -------------------------- 73
`General Plastic Indus. Co. v. Canon Kabushiki Kaisha,
`IPR2016-01357, Paper 19 (P.T.A.B. Sept. 6, 2017) --------------------------- 76
`Global Tel*Link Corp. v. HLFIP Holding, Inc.,
`IPR2021-00444, Paper 14 (P.T.A.B. Jul. 22, 2021) ---------------------------- 73
`Google LLC v. Jawbone Innovations, LLC,
`IPR2022-00630, Paper 10 (P.T.A.B. Sept. 13, 2022) -------------------------- 75
`Google LLC v. Jawbone Innovations LLC,
`IPR2022-00630, Paper 13 (P.T.A.B. October 28, 2022) ---------------------- 74
`
`Google LLC v. Jawbone Innovations, LLC,
`IPR2022-01124, Paper 11 (P.T.A.B. Jan. 3, 2023) ---------------------------- 72
`
`-v-
`
`
`
`TABLE OF AUTHORITIES
`(Cont’d)
`
`Page No.
`
`Jawbone Innovations, LLC v. LG Electronics Inc.,
`No. 2:23-cv-00078-JRG-RSP (E.D. Tex.) -------------------------------------- 72
`KSR Int’l Co. v. Teleflex Inc.,
`550 U.S. 398 (2007) ---------------------------------------------- 45, 46, 49, 68, 71
`Leapfrog Enters. v. Fisher-Price, Inc.,
`485 F.3d 1157 (Fed. Cir. 2007) --------------------------------------------------- 71
`Mercedes-Benz USA, LLC v. Carucel Invs. L.P.,
`IPR2019-01404, Paper 12 (Jan. 22, 2020) --------------------------------------- 77
`NetNut Ltd. v. Bright Data Ltd.,
`IPR2021-00465, Paper 11 (P.T.A.B. Aug. 12, 2021) -------------------------- 76
`Newell Cos. v. Kenney Mfg. Co.,
`864 F.2d 757 (Fed. Cir. 1988) ----------------------------------------------------- 71
`Nidec Motor Corp. v. Zhongshan Broad Ocean Motor Co. Ltd.,
`868 F.3d 1013 (Fed. Cir. 2017) --------------------------------------------------- 24
`Samsung Elecs. Am. Inc. v. Snik LLC,
`IPR2020-01428, Paper 10 (P.T.A.B. Mar. 9, 2021) ---------------------------- 74
`Sand Revolution II, LLC v. Cont’l Intermodal Grp.-Trucking LLC,
`IPR2019-01393, Paper 24 (P.T.A.B. June 16, 2020 --------------------------- 74
`Toshiba Am. Info. Sys., Inc. v. Walletex Microelecs. Ltd.,
`IPR2018-01538, Paper 11 (Mar. 5, 2019) --------------------------------------- 77
`Uniloc 2017 LLC v. Samsung Elecs. Am., Inc.,
`No. 2:19-cv-00259, 2020 WL 1433960 (E.D. Tex. Mar. 24, 2020) --------- 72
`Unwired Planet, LLC v. Google Inc.,
`841 F.3d 995 (Fed. Cir. 2016) ----------------------------------------------------- 23
`Vivid Techs., Inc. v. Am. Sci. & Eng’g, Inc.,
`200 F.3d 795 (Fed. Cir. 1999) ----------------------------------------------------- 24
`
`
`
`vi
`
`
`
`TABLE OF AUTHORITIES
`(Cont’d)
`
`Page No.
`
`Statutes and Rules:
`
`35 U.S.C. §325(d) ---------------------------------------------------------------------- 77
`
`
`
`vii
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`Exhibit No.
`
`TABLE OF EXHIBITS
`Description
`
`1001
`
`1002
`
`1003
`
`1004
`
`1005
`
`1006
`
`1007
`
`1008
`
`1009
`
`1010
`
`1011
`
`1012
`
`U.S. Patent No. 11,122,357 (“the ’357 patent”)
`
`Declaration of Richard M. Stern, Ph.D.
`
`Excerpts of MICROPHONE ARRAYS: SIGNAL PROCESSING
`TECHNIQUES AND APPLICATIONS (Michael Brandstein & Darren
`Ward eds., Springer-Verlag 2001) (“Brandstein”)
`
`Sharon Gannot et al., Signal Enhancement Using Beamforming
`and Nonstationarity with Applications to Speech, vol. 49, no. 8
`IEEE TRANSACTIONS ON SIGNAL PROCESSING, 1614 (Aug. 2001)
`(“Gannot”)
`
`Lloyd Griffiths & Charles Jim, An Alternative Approach to
`Linearly Constrained Adaptive Beamforming, vol. AP-30, no. 1
`IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, 27 (Jan.
`1982) (“Griffiths-Jim”)
`
`Iain A. McCowan et al., Near-Field Adaptive Beamformer for
`Robust Speech Recognition, vol. 12, no. 1 DIGITAL SIGNAL
`PROCESSING, 87 (Jan. 2002) (“McCowan”)
`
`U.S. Patent No. 5,651,071 (“Lindemann”)
`
`U.S. Patent No. 5,627,799 (“Hoshuyama”)
`
`U.S. Patent Publication No. 2003/0128848 (“Burnett ’848”)
`
`Excerpts from the ’357 patent’s file history
`
`U.S. Provisional Patent Application No. 61/045,377
`
`Curriculum Vitae of Richard M. Stern, Ph.D.
`
`Table of Exhibits, Page 1
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`Exhibit No.
`
`Description
`
`1013
`
`Declaration of Carol S. Peterson
`
`
`
`
`
`Table of Exhibits, Page 2
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`Petitioner LG Electronics Inc. (“LGE” or “Petitioner”) requests inter partes
`
`review (“IPR”) of Claims 1-20 of U.S. Patent 11,122,357, which Jawbone
`
`Innovations, LLC (“Patent Owner” or “PO”) purportedly owns.
`
`INTRODUCTION
`The challenged claims recite devices that process audio signals from
`
`microphones to reduce noise. The claimed devices comprise two “virtual”
`
`microphones, each formed by combining signals from two physical microphones.
`
`The two virtual microphones must have substantially similar responses to noise and
`
`substantially dissimilar responses to speech. But the claims do not require that the
`
`device do anything with the virtual microphones. Instead, the claims merely recite
`
`that the device must include a signal processor that performs generic signal-
`
`processing operations like filtering the physical-microphone signals, summing the
`
`physical-microphone signals, and applying a transfer function. The claims recite
`
`these conventional signal-processing operations only at a high level. For example,
`
`the claims do not elaborate on the filter to apply; they require only that some
`
`“filtering” of the physical-microphone signals occur.
`
`The two virtual microphones may be created by the recited filtering and
`
`summing of the physical-microphone signals. But as the ’357 patent concedes, that
`
`was a “common” technique for creating virtual microphones known to those skilled
`
`in the art. (Ex. 1001, 8:55-60.) The generic language of the claims contrasts with the
`
`1
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`’357 patent specification, which identifies specific formulas defining the two virtual
`
`microphones that are the basis for the purported innovation. (Compare, e.g., id.,
`
`claim 1 with id., 11:6-16, 12:20.)
`
`Untethered from the specification’s formulas, the claims’ recitation of generic
`
`signal-processing concepts encompasses prior art describing the Generalized
`
`Sidelobe Canceler (“GSC”), a fundamental noise-reduction technique introduced in
`
`the 1980s. Broadly applicable to many signal-processing applications, the GSC
`
`involves filtering and summing the signals from at least two sensors in different
`
`ways to produce two virtual sensors, one that captures the target signal plus noise
`
`and another that captures just noise. Subtracting the noise signal from the target-plus
`
`noise signal cancels out the noise and yields a cleaner output signal.
`
`Years before the ’357 patent’s earliest possible priority date, a widely used
`
`reference book, MICROPHONE ARRAYS (Springer-Verlag 2001) (“Brandstein”),
`
`explained that it was common to use the GSC with a microphone array to reduce
`
`noise
`
`in speech-signal processing. Contemporaneously with Brandstein’s
`
`publication, Sharon Gannot and other researchers published in IEEE’s Transactions
`
`on Signal Processing an article titled Signal Enhancement Using Beamforming and
`
`Nonstationarity with Applications to Speech (“Gannot”), describing a generalized
`
`version of the GSC technique that would make it even more robust by handling
`
`arbitrary linear transfer functions. As its title indicates, Gannot likewise
`
`2
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`contemplated reducing noise in speech applications. Together, Brandstein and
`
`Gannot disclose all the limitations of the ’357 patent’s claims and render all the
`
`claims obvious.
`
`Because they cover GSC techniques published years before the earliest
`
`priority date, claims 1-20 are unpatentable. The Board should cancel those claims.
`
`I.
`
`BACKGROUND1
`When Patent Owner filed the ’357 patent’s priority applications in 2007,
`
`techniques for reducing noise in signals had been known for decades. One prominent
`
`technique was the Generalized Sidelobe Canceler, or GSC. In a nod to its inventors,
`
`the GSC is sometimes also called the Griffiths-Jim beamformer. (Ex. 1002 ¶30.)
`
`A. Griffiths and Jim Publish Their Seminal GSC Article in 1982.
`In 1982, Lloyd Griffiths and Charles Jim published a paper describing “a
`
`simple time-varying beamformer which can be used to combine the outputs of an
`
`array of sensors.” (Ex. 1005, 27). The beamformer’s purpose was “to minimize the
`
`
`1 Sections I-VII of this petition are substantively identical to the corresponding
`
`sections of the petition in Amazon.com, Inc. v. Jawbone Innovations, LLC, IPR2023-
`
`00251, Paper 1 (PTAB Nov. 21, 2022), which Petitioner seeks to join pursuant to
`
`the motion for joinder and consolidation filed concurrently herewith.
`
`3
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`effects of noise and interference at the array output” while capturing the target signal.
`
`(Id.)
`
`Griffiths and Jim called their beamformer a “generalized sidelobe canceling”
`
`structure. (Id., 29.) Illustrated in Figure 4 of their paper, the signal processor had two
`
`main substructures: the top branch was a “conventional beamformer” designed to
`
`capture the target signal plus noise, and the bottom branch was the “sidelobe
`
`canceling path” that captures only noise so that the noise could be subtracted or
`
`canceled out:
`
`
`(Id., 29–30.2) In the top branch, the outputs of the array sensors were combined to
`
`form a conventional beamformer, which Petitioner calls the first virtual sensor. The
`
`sensor outputs were combined by multiplying the sensor output signals by factors
`
`
`2 Figures have been annotated with color throughout.
`
`4
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`called “weights” (𝑤𝑐1, … , 𝑤cM in the paper, and sometimes also called “gains”) and
`the paper’s equations, the output of the first virtual sensor was denoted 𝑦(cid:3030)(cid:4594)(cid:4666)𝑘(cid:4667) (Ex.
`
`further filtering and summing the weighted sensor signals. (Id.; Ex. 1002 ¶35.) In
`
`1005, 30.) This output contains the target signal plus noise. (Ex. 1002 ¶35.)
`
`“The lower path in Fig. 4 is the sidelobe canceling path” (Ex. 1005, 30), which
`
`Petitioner calls the second virtual sensor. Like the first virtual sensor, the second
`
`virtual sensor is formed from a combination of the outputs of the array sensors, but
`
`the combination differs from the combination used for the first virtual sensor. The
`
`from the lower path.” (Id.) The array sensor outputs were combined by filtering and
`
`summing (including by blocking the desired signal), and the output of the second
`
`lower path includes a blocking matrix designed to “block the desired signal 𝑠(cid:4666)𝑘(cid:4667)
`virtual sensor was denoted in the paper as 𝑦(cid:3002)(cid:4666)𝑘(cid:4667). (See id.) The second virtual
`sensor’s output “𝑦(cid:3002)(cid:4666)𝑘(cid:4667) contains no desired signal terms” and instead “contains only
`The overall output of the GSC, 𝑦(cid:4666)𝑘(cid:4667), was produced by subtracting the noise
`
`noise and interference terms.” (Id.)
`
`only output of the second virtual sensor from the target-plus-noise output of the first
`
`virtual sensor:
`
`
`
`5
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`(Id.) The result was a cleaned-up signal that reduced noise without distorting the
`
`desired signal. (Ex. 1002 ¶40; Ex. 1005, 30 (output due to desired signal satisfies the
`
`constraint defined by paper’s equation 9, which defines a constraint for “zero
`
`distortion” (p. 28)).)
`
`B. Over the Next Two Decades, the GSC was Used in Microphone
`Arrays to Reduce Noise.
`In the twenty years following Griffiths and Jim’s article, the GSC was used in
`
`many signal-processing applications, including with microphone arrays to reduce
`
`noise
`
`in speech applications. For example, U.S. Patent No. 5,651,071
`
`(“Lindemann”), filed in 1993, cites the article and explains that using a Griffiths-Jim
`
`beamformer “to improve signal-to-noise ratio for hearing aids” was known. (Ex.
`
`1007, 1:40-46, 12:12–14.3) As another example, Griffiths-Jim is the first non-patent
`
`reference cited in U.S. Patent No. 5,627,799 (“Hoshuyama”), filed in 1995, which
`
`relates to “interference cancelers, and more particularly to a generalized sidelobe
`
`canceler, or adaptive beamformer for an array of sensors such as microphones[.]”
`
`(Ex. 1008, 1:8-11.) Describing what was prior art even then, Hoshuyama explains
`
`one way the GSC had been used with microphone arrays:
`
`According to a prior art microphone array, signals detected by an array
`of microphones are lowpass filtered and summed together to detect a
`target signal that arrives in a particular direction. The adaptive
`microphone array beamformer is one form of the generalized sidelobe
`
`3 Patent citations are in column:line format.
`
`6
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`canceler as described in an article “An alternative Approach to Linearly
`Constrained Adaptive Beamforming”, Lloyd J. Griffiths and Charles
`W. Jim, the IEEE Transactions on Antenna and Propagation, Vol. AP-
`30, No. 1, January 1982, pages 27-34.
`(Id., 1:17-26.)
`
`C.
`
`Published in 2001, Brandstein Illustrates How to Use a GSC with
`a Microphone Array to Reduce Noise.
`In 2001, MICROPHONE ARRAYS: SIGNAL PROCESSING TECHNIQUES AND
`
`APPLICATIONS published. (Ex. 1003 (“Brandstein”).) The editors’ goal was to
`
`provide “a single complete reference on microphone arrays.” (Id., Preface.) The
`
`book quickly became a standard reference for those in the field of audio-signal
`
`processing. (Ex. 1002 ¶45.)
`
`At the outset of the chapter on robust adaptive beamforming, Brandstein
`
`explains that “[a]pplications of beamforming include microphone arrays for speech
`
`enhancement.” (Ex. 1003, 87 (original page numbering).) “The goal of speech
`
`enhancement is to remove undesirable signals such as noise and reverberation.” (Id.)
`
`Brandstein further explains that, among various known adaptive beamformers, “the
`
`Griffiths-Jim beamformer (GJBF), or the generalized sidelobe canceler, is most
`
`widely known.” (Id., 88 (internal citation omitted).) “Figure 5.1 depicts the structure
`
`of the GJBF.” (Id.)
`
`7
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`
`
`(Id., 89, Fig. 5.1.) As shown, the signals of at least two physical microphones, 𝑥0 (𝑘)
`and 𝑥1 (𝑘), are combined in the top branch by filtering and summing the signals to
`
`form a fixed beamformer—a first virtual microphone. (Ex. 1002 ¶48.) The first
`
`virtual microphone captures the target speech signal plus noise. (Id.)
`
`In the bottom branch, the signals of the two physical microphones are
`
`combined by filtering and summing the signals in a different way to form a second
`
`virtual microphone. (Id. ¶49.) The second virtual microphone includes a blocking
`
`matrix (BM). (Id.) “[T]he BM forms a null in the look direction so that the target
`
`signal is suppressed and all other signals are passed through.” (Ex. 1003, 88.) “The
`
`8
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`BM was named after its function, which is to block the target signal.” (Id.) As a
`
`result, the second virtual microphone captures noise only. (Ex. 1002 ¶49.)
`
`The overall output is the target-plus-noise output of the first virtual
`
`microphone minus the noise-only output of the second virtual microphone. (Id. ¶50.)
`
`The result is that, “in the subtracter output 𝑦(𝑘), the target signal is enhanced and
`
`undesirable signals such as ambient noise and interferences are suppressed.” (Ex.
`
`1003, 88-89.)
`
`The two virtual microphones have very different responses to the target
`
`speech signal: the first virtual microphone is designed to capture the target signal,
`
`while the second virtual microphone is designed to block it. (Ex. 1002 ¶51.) On the
`
`other hand, they have similar responses to noise so that in the final subtraction output
`
`the noise is removed. (Id.) This is illustrated in Figure 5.2, which shows the
`
`directivity pattern for the final output of an example Griffiths-Jim beamformer:
`
`(Ex. 1003, 89, Fig. 5.2.) The horizontal axis of the graph shows direction of arrival
`
`measured in degrees relative to the microphone array: the target signal is shown at 0
`
`
`
`9
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`degrees, while the noise signal is shown at approximately 45 degrees. (Ex. 1002
`
`¶51.) The vertical axis of the graph shows gain in decibels: zero gain corresponds to
`
`no change in sound pressure or signal power, a positive gain corresponds to an
`
`increase in signal power, and a negative gain corresponds to a decrease in signal
`
`power. (Id.)
`
`As highlighted in green, the target signal is reproduced faithfully with
`
`essentially zero gain, reflecting that subtracting the second virtual microphone’s
`
`noise-only output from the first virtual microphone’s target-plus-noise output will
`
`produce the target signal in the direction of the target. (Id. ¶52.) On the other hand,
`
`as highlighted in red, Figure 5.2 shows a highly negative gain in the direction of the
`
`interference signal, reflecting that subtracting the two virtual microphones’ outputs
`
`cancels the noise. (Id. ¶53; see also Ex. 1003, 90 (“In the direction of the target
`
`signal, almost constant gains close to 0 dB are obtained over a wide range of
`
`frequencies. On the contrary, in the direction of the interference, a deep null is
`
`formed.”).)
`
`D. Contemporaneously, Gannot Taught Adapting the GSC to Handle
`Arbitrary Transfer Functions.
`In August 2001, IEEE’s Transactions on Signal Processing publication (vol.
`
`49, no. 8) included an article titled Signal Enhancement Using Beamforming and
`
`Nonstationarity with Applications to Speech by Sharon Gannot, David Burshtein,
`
`and Ehud Weinstein. (Ex. 1004 (“Gannot”).) As its title indicates, the article
`
`10
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`contemplates speech enhancement through beamforming. (Id.) Specifically, the
`
`article considers a sensor array “where arbitrary transfer functions (TFs) relate the
`
`source signal and the sensors.” (Id., 1614 (Abstract).) As an audio signal travels from
`
`its source to a microphone, the signal may change, such that the signal received at
`
`the microphone is not exactly the same as the signal when it originated from the
`
`source. (Ex. 1002 ¶55.) The acoustic path from the source to the microphone can be
`
`thought of as a system that brings about this change, and the operation of the acoustic
`
`path on the signal can be represented mathematically by a transfer function. (Id.)
`
`Gannot notes that the generalized sidelobe canceler (GSC) works well when
`
`the acoustic paths’ transfer functions satisfy certain criteria, such as when the signals
`
`received at the sensors “are simple delayed versions of the source signal.” (Ex. 1004,
`
`1614 (Abstract).) But the original Griffiths-Jim GSC may suppress interference
`
`poorly “in complicated acoustic environments, where arbitrary TFs [transfer
`
`functions] may be encountered.” (Id.) Gannot thus proposes a GSC solution adapted
`
`to handle arbitrary transfer functions. (Id.)
`
`11
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`Gannot’s Figure 3 shows the proposed GSC structure, and Figure 4 summa-
`
`rizes the algorithm:
`
`The blocking matrix ℋ† is used to create noise reference signals 𝑈(cid:3040) that apply the
`linear transfer functions 𝐴(cid:3040) of the acoustical paths:
`
`(Ex. 1004, 1618-20.) By incorporating these terms, Gannot’s more-general GSC
`
`accounts for arbitrary transfer functions. In particular, by using the ratio of the
`
`transfer function for microphone m to the transfer function for microphone
`
`12
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`1, Gannot’s algorithm applies a transfer function for an acoustical path between the
`
`two microphones.4 (Ex. 1002 ¶61.)
`
`E. Other Concepts in the ’357 Patent Were Well Known.
`1.
`Filtering and Summing in the Time Domain Were Well
`Known.
`In the microphone-array context, the GSC involves filtering and summing
`
`physical-microphone signals to create virtual microphones. It has been known since
`
`the outset that the GSC can be implemented in the time domain, a term that refers to
`
`analyzing signals as a function of time. (Ex. 1002 ¶35.) Both the original Griffiths-
`
`Jim paper and Brandstein show the sensor signals as functions of time. (Id. ¶¶35,
`
`48.) Indeed, Lindemann discloses that using the GSC in a microphone array for
`
`hearing aids was a “time domain approach.” (Ex. 1007, 1:51-52.)
`
`
`4 The inventor named on the ’357 patent had another application publish as U.S.
`
`2003/0128848 (“Burnett ’848”), which is prior art to the ’357 patent. (Ex. 1009.)
`
`Burnett ’848 also discloses using a ratio of transfer functions representing the
`
`transfer function of an acoustical path between microphones. (Ex. 1002 ¶¶71-75.)
`
`The ’357 patent’s discussion of transfer functions essentially repeats Burnett ’848’s
`
`disclosure, which confirms that this feature was known in the art. (Id. ¶83.)
`
`13
`
`
`
`Petition for Inter Partes Review of U.S. Patent No. 11,122,357
`
`2.
`
`Delaying Signals Based on Geometry to Adjust for
`Differences in Arrival Times Was Well Known.
`Adaptive beamforming relies on the spatial geometry of the array and the
`
`signal’s direction of arrival. (Ex. 1002 ¶62; Ex. 1003, 87.) When the source of the
`
`target signal is close enough to the array, instead of treating the signal’s wavefront
`
`as a plane, it may be useful to account for the spherical geometry of the wavefront—
`
`i.e., to use a near-field design. (See Ex. 1002 ¶¶63-64.) Near-field designs were well
`
`known before the ’357 patent and use basic calculations to determine the