`571-272-7822
`
`Paper 58
`Entered: November 29, 2022
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`APPLE INC.,
`Petitioner,
`v.
`ANDREA ELECTRONICS CORP.,
`Patent Owner.
`
`IPR2017-00626
`Patent 6,363,345 B1
`
`
`
`
`
`
`
`
`
`Before MICHAEL R. ZECHER, JEREMY M. PLENZLER, and
`MIRIAM L. QUINN, Administrative Patent Judges.
`PLENZLER, Administrative Patent Judge.
`
`
`
`JUDGMENT
`Final Written Decision on Remand
`Determining Challenged Claims 6–9 Are Unpatentable
`35 U.S.C. §§ 144, 318(a)
`
`INTRODUCTION
`I.
`A. Background and Summary
`Apple Inc. (“Petitioner”) requested inter partes review of claims 1–25
`and 38–47 of U.S. Patent No. 6,363,345 B1 (Ex. 1001, “the ’345 patent”).
`
`
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`Paper 1 (“Pet.”). We issued a Decision instituting inter partes review.
`Paper 7 (“Inst. Dec.”).
`After institution of trial, Andrea Electronics Corp. (“Patent Owner”)
`filed a Patent Owner Response (Paper 11, “PO Resp.”), to which Petitioner
`filed a Reply (Paper 18, “Pet. Reply”). An oral argument was held on April
`25, 2018. A transcript of the oral argument is included in the record. Paper
`25.
`
`Our Final Written Decision was issued on July 12, 2018. Paper 28
`(“Original Decision” or “Original Dec.”). The Original Decision determined
`that Petitioner had established unpatentability of claims 1–3, 12–25, 38, and
`47 of the ’345 patent, but had not established unpatentability of claims 4–11
`and 39–46 of the ’345 patent. Original Decision 24.
`On appeal, the U.S. Court of Appeals for the Federal Circuit vacated
`our Original Decision only as to claims 6–9 of the ’345 patent and remanded
`the case for further proceedings. Apple Inc. v. Andrea Elecs. Corp., 949
`F.3d 697 (Fed. Cir. 2020). After conferring with the parties, we permitted
`additional briefing addressing the issues on remand from the Federal Circuit.
`Paper 36. Petitioner and Patent Owner simultaneously filed opening briefs
`(Paper 41 (“Pet. Remand Br.”); Paper 42 (“PO Remand Br.”)), followed by
`simultaneously filed Reply Briefs (Paper 44 (“Pet. Remand Reply”); Paper
`43 (“PO Remand Reply”)). After further conferring with the parties, we
`authorized an additional round of briefing (Paper 45), which was filed to
`address claim 9 of the ’345 patent. Paper 46 (“Pet. Remand Sur-Reply”);
`Paper 49 (“PO Remand Sur-Reply”). In view of the remand, we issued a
`Final Written Decision on Remand only as to the patentability of challenged
`claims 6–9, determining those claims unpatentable. Paper 50 (“Remand
`Decision” or “Remand Dec.”).
`
`2
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`On appeal from our Remand Decision, “[Patent Owner] d[id] not
`dispute that Martin discloses all of the limitations of claims 6 through 9,
`with the exception of the step in claim 9 ‘wherein said future minimum
`value is set to a current magnitude value periodically.’” Andrea Elecs. Corp.
`v. Apple Inc., No. 2021-1248, 2022 WL 1197341, at *2 (Fed. Cir. Apr. 22,
`2022). The Federal Circuit “affirm[ed] [our] decision finding that Martin
`discloses the limitations of claim 9.” Id. at *5. The only issue remaining
`before us is whether there is sufficient motivation to combine the teachings
`of Hirsch and Martin with respect to claims 6–9. Id.
`For the reasons discussed below, we determine that Petitioner has
`shown by a preponderance of the evidence that one skilled in the art would
`have been motivated to combine the teachings of Hirsch and Martin with
`respect to claims 6–9 and, therefore, Petitioner has shown that these claims
`are unpatentable.
`B. Prior Art and Asserted Grounds Relevant to Remand
`Petitioner asserts that claims 6–9 would have been unpatentable based
`on the following ground, which is the ground remaining for consideration on
`remand:
`Claim(s) Challenged
`6–9
`
`Reference(s)/Basis
`Hirsch1, Martin2
`
`35 U.S.C. §
`103
`
`C. CAFC Remand
`On appeal, the Federal Circuit held that our Remand Decision erred
`by failing to properly analyze the motivation to combine the teachings of
`
`
`1 H.G. Hirsch & C. Ehrlicher, “Noise Estimation Techniques for Robust
`Speech Recognition,” IEEE 1995 (Ex. 1005, “Hirsch”).
`2 Ranier Martin, “An Efficient Algorithm to Estimate the Instantaneous SNR
`of Speech Signals,” Eurospeech 1993 (Ex. 1006, “Martin”).
`
`3
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`Hirsch with those of Martin with respect to claims 6–9. Andrea, 2022 WL
`1197341, at *1, 5.
`
`II. ANALYSIS
`A. Scope of Remand
`There is no remaining dispute as to whether the combination of Hirsch
`and Martin teaches the limitations of claims 6–9. Hirsch teaches the noise
`cancelling apparatus recited in claim 1, including the threshold detector
`recited in that claim, as well as the additional details of that threshold
`detector recited in claims 4 and 5 depending from claim 1. See Original
`Dec. 6–7, 24; Remand Dec. 6. Martin teaches a threshold detector having
`the additional features recited in claims 6–9, which depend from claim 5.
`See Remand Dec. 6–17.
`In its remand decision, the Federal Circuit stated that “Hirsch does
`refer to Martin as a known approach to avoid the problem of speech pause
`detection to estimate noise and, based on that, it was reasonable for the
`Board to find that a skilled artisan would have considered Martin’s teachings
`generally when reviewing Hirsch.” Andrea, 2022 WL 1197341, at *5. The
`Federal Circuit remanded for us to “address[] why a skilled artisan would
`have specifically incorporated Martin’s noise power estimation algorithm
`into Hirsch’s spectral subtraction system or threshold calculation” and to
`“address[] any of the specific motivation issues contested by [Patent
`Owner].” Id. Accordingly, the Federal Circuit remanded solely “for further
`findings on the motivation to combine Hirsch and Martin with respect to
`claims 6–9.” Id. That is, the scope of the remand is limited and requires us
`to make specific fact findings related to whether/why Martin’s noise power
`estimation algorithm would have been incorporated into Hirsch.
`
`4
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`
`B. Petitioner’s Combination
`With respect to claims 6–9, Petitioner’s combination of Martin’s
`teachings with those of Hirsch is “to incorporate Martin’s technique for
`estimating the noise floor of a signal within Hirsch’s adaptive threshold
`calculation to improve operation of Hirsch’s system in non-stationary noise
`settings.” Pet. 35 (citing Ex. 1003 ¶¶ 129–132).
`At this stage of the proceeding, there is no dispute that Hirsch
`specifically identifies the problem of speech pause detection to estimate
`noise, and specifically identifies Martin as a solution to that problem,
`particularly in non-stationary noise settings. Petitioner contends that
`because “Hirsch . . . did not test its system with non-stationary noise signals”
`and “Martin . . . indicates that its noise floor estimation technique provided
`good results in both stationary and non-stationary noise settings,” one skilled
`in the art would have been “motivated . . . to alter Hirsch’s scheme to use the
`Martin noise floor estimation technique.” Pet. 35–36 (citing Ex. 1005, 154;
`Ex. 1006, 1095–1096; Ex. 1003 ¶¶ 128–129, 132). We agree.
`The cited portions of Martin support Petitioner’s position. See
`Ex. 1006, 1095–1096 (discussing testing of the algorithm in non-stationary
`noise settings and concluding that the arrangement “provides a
`computational[ly] inexpensive and effective mean to cope with th[e]
`problem”). And there is no dispute that Hirsch is silent as to the
`effectiveness of its technique in non-stationary noise environments. See
`Pet. 35 (“Hirsch explains its noise estimation scheme worked well for
`stationary noise signals, but did not test its system with non-stationary noise
`signals.”) (citing Ex. 1005, 154); PO Resp. 38 (“Hirsch includes no
`explanation as to why Hirsch elected to test his techniques only in stationary
`noise environments.”) (citing Ex. 2002 ¶ 87); PO Remand Br. 5 (“Nothing in
`
`5
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`Hirsch discusses how its algorithm might perform in non-stationary noise
`environments.”) (citing Ex. 2002 ¶ 88). Hirsch, however, acknowledges that
`“detection of speech pauses to evaluate segments of pure noise . . . is a
`difficult task especially if the background noise is not stationary or the
`signal-to-noise ratio (SNR) is low.” Ex. 1005, 153.
`Petitioner presents sufficient reasoning to incorporate Martin’s
`algorithm into Hirsch’s system because Hirsch, itself, notes the difficulty of
`addressing non-stationary noise environments, while leaving unaddressed
`whether its system would be effective in such environments. Martin, on the
`other hand, is expressly referenced in Hirsch as a known solution to the
`problem identified in Hirsch, and provides a noise floor algorithm that
`provides good results in non-stationary noise environments.
`Patent Owner disputes various aspects of the rationale provided by
`Petitioner, which we address in more detail below. Based on the record
`before us, however, the preponderance of the evidence supports Petitioner’s
`position that one skilled in the art would have appreciated that Martin
`provides a proven way to deal with non-stationary noise environments that
`one skilled in the art would have been motivated to substitute into Hirsch’s
`system based on Hirsch’s identification of this known problem and Martin as
`a known solution to this problem.
`
`6
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`
`C. Patent Owner’s Rebuttal
`1. Whether Hirsch shows a need for improvement in non-stationary noise
`environments and whether Martin provides such improvement
`Patent Owner contends that “there is no basis that a skilled artisan
`would have concluded that Hirsch’s noise estimation techniques needed to
`be improved in non-stationary noise environments.” PO Resp. 37. Patent
`Owner acknowledges that “Hirsch appears to have only tested its techniques
`in stationary noise settings, but [contends that] there is no evidence
`suggesting that Hirsch’s techniques would not work well, or would need to
`be improved, in non-stationary noise environments.” Id. at 37–38. Patent
`Owner contends that “Hirsch does not disclose that its techniques would
`perform worse or not as well in non-stationary noise environments and
`Hirsch includes no explanation as to why Hirsch elected to test his
`techniques only in stationary noise environments.” Id. at 38. We agree that
`Hirsch is silent on whether its system needed improvement in non-stationary
`noise settings because Hirsch says nothings about its performance in non-
`stationary noise settings. Compared to Hirsch’s unproven system, Martin’s
`system is an improvement in non-stationary noise settings by virtue of being
`a system tested and shown to be effective in such settings, as explained
`below.
`There is no dispute that Martin provides an “approach[] known to
`avoid the problem of speech pause detection and to estimate the noise
`characteristics just from a past segment of noisy speech,” as acknowledged
`by Hirsch, itself. Ex. 1005, 153; see also PO Resp. 35. Martin’s “algorithm
`is capable to track non[-]stationary noise signals and has a low
`computational complexity.” Ex. 1006, 1093. Martin tested its algorithm
`with non-stationary noise and concluded that “[v]arying noise levels have a
`
`7
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`significant impact on the performance of many speech processing
`algorithms” and “[t]he algorithm proposed in [Martin] provides a
`computational[ly] inexpensive and effective mean to cope with this
`problem.” Id. at 1095–1096.
`Patent Owner disagrees with the effectiveness of Martin’s algorithm.
`Specifically, Patent Owner contends that “Martin does not report that the use
`of his algorithm in his experiments produced good results in cases of non-
`stationary noise environments.” PO Resp. 37 (citing Ex. 1006, 1094); see
`also id. 39 (citing Ex. 2002 ¶ 87). Patent Owner additionally contends that
`“Martin also reports less-than-satisfactory results when his algorithm was
`applied to a spectral subtraction scheme.” Id. at 39 (citing Ex. 2002 ¶ 87).
`This is inconsistent with the express teachings of Martin discussed above.
`We credit Dr. Hochwald’s testimony in this regard, over that of Dr. Douglas,
`because Dr. Hochwald’s testimony is consistent with the disclosure of
`Martin. See, e.g., Ex. 1003 ¶¶ 128–129 (“Martin explains that the noise
`floor algorithm is particularly suited to identifying noise in a frequency bin
`during active speech [Martin, pg. 1094]” and “Martin’s algorithm has some
`advantages in that it had been tested to work well with non-stationary
`noise.”).
`Patent Owner and Dr. Douglas focus on inaccuracies discussed in
`Martin during certain circumstances, but fail to address the expressly stated
`overall effectiveness of Martin’s algorithm in non-stationary noise settings
`as discussed above. See, e.g., Ex. 1006, 1095–1096. Dr. Douglas, for
`example, testifies that “Martin concluded that . . . his algorithm did not
`produce good results during periods of nonspeech activity,” with little
`explanation, other than asserting that “Martin states on page 1095 that his
`‘estimate is biased when no speech is present’ and states on page 1096 that
`
`8
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`his ‘algorithm is accurate for medium to high SNR conditions, but
`necessarily biased when no speech is present.’” Ex. 2002 ¶ 87.
`Although Martin explains that “[t]he algorithm is accurate for medium
`to high SNR conditions but necessarily biased when no speech is present,”
`that bias is described as “slightly biased” in the preceding paragraph, which
`Patent Owner acknowledges (PO Resp. 39). See Ex. 1006, 1096. We do not
`agree with the characterization of Martin’s algorithm as being ineffective for
`non-stationary noise situations when applied to a spectral subtraction scheme
`because such a determination would ignore the express characterization in
`Martin that its algorithm is, in fact, “a computational[ly] inexpensive and
`effective mean to cope with th[e] problem.” Id.
`2. Whether Hirsch disparages Martin because of the difference in length of
`past segments of noise speech and alleged significant difference in time
`requirements between Martin and Hirsch
`Patent Owner contends that “Hirsch immediately disparages
`techniques like Martin’s, explaining that ‘[t]he disadvantage of most
`approaches is the need of relatively long past segments of noisy speech.’”
`PO Resp. 35 (quoting Ex. 1005, 153). Patent Owner contends that “[t]he
`relevant time frames of noisy speech segments required by the systems
`described in Hirsch and Martin confirm Hirsch’s criticism of Martin . . . a
`significant difference of more than 50%.” Id. (citing Ex. 2002 ¶¶ 83–84).
`We disagree with Patent Owner that one skilled in the art would not have
`looked to Martin’s teachings regarding estimating the noise floor of a signal
`in non-stationary noise settings because Hirsch allegedly disparages Martin.
`See id. at 35–36.
`Hirsch states that “[s]ome approaches are known to avoid the problem
`of speech pause detection and to estimate the noise characteristics just from
`
`9
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`a past segment of noisy speech.” Ex. 1005, 153 (citing Martin and two other
`references). Hirsch further explains that “[t]he disadvantage of most
`approaches is the need of relatively long past segments of noisy speech”
`without specifying the time period that constitutes “a relatively long past
`segment” or if Martin is one “of most approaches.” Id. (emphasis added).
`We read Hirsch as simply describing alternate approaches. See In re
`Fulton, 391 F.3d 1195, 1201 (Fed. Cir. 2004) (The “mere disclosure of
`alternative designs does not teach away.”). “A reference that ‘merely
`expresses a general preference for an alternative invention but does not
`criticize, discredit, or otherwise discourage investigation into’ the claimed
`invention does not teach away.” Meiressone v. Google, Inc., 849 F.3d 1379,
`1382−83 (Fed. Cir. 2017) (citing Galderma Labs., L.P. v. Tolmar, Inc., 737
`F.3d 731, 738 (Fed. Cir. 2013) and finding that a reference that provided a
`feature that may have been inferior to another does not teach away from its
`use because it does not amount to criticizing or promoting abandonment of
`that feature). Martin’s algorithm requires 625 milliseconds to estimate
`noise, whereas Hirsch needs past noisy segments of about 400 milliseconds.
`The parties proffer competing testimony as to whether the 225 millisecond
`difference is significant. See Ex. 1003 ¶ 129; Ex. 2002 ¶ 85.
`Dr. Hochwald testifies that “[t]o a person knowledgeable in the arts,
`the difference between 0.4 seconds and 0.625 seconds is not very significant
`in capturing speech data” without further evidence or explanation to support
`that conclusion. Ex. 1003 ¶ 129. Dr. Douglas testifies without further
`evidence or explanation that “[i]n [his] opinion a skilled artisan would have
`thought that in the context of Hirsch’s and Martin’s disclosures the
`difference between 625 milliseconds and 400 milliseconds is significant.”
`Ex. 2002 ¶ 85. Neither testimony is helpful because those statements
`
`10
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`present conclusory statements as to the significance of the 225 milliseconds
`difference. However, we can discern from the references themselves that
`Hirsch does not teach away. See Belden Inc. v. Berk-Tek LLC, 805 F.3d
`1064, 1079 (Fed. Cir. 2015) (“Board members, because of expertise, may
`more often find it easier to understand and soundly explain the teachings and
`suggestions of prior art without expert assistance.”).
`As explained above, the 400 millisecond scenario in Hirsch is
`discussed in connection with evaluation of stationary noise signals. The
`record is silent about the effectiveness of Hirsch’s 400 millisecond sample
`size in non-stationary noise settings. Martin, on the other hand, specifically
`tailors its window length to non-stationary noise settings, explaining that
`“[t]he window length . . . must be large enough to bridge any peak of speech
`activity, but short enough to follow non[-]stationary noise variations” and
`“[e]xperiments . . . have shown that a window length of 0.625 s is a good
`value.” Ex. 1006, 1094. Thus, we understand Martin’s 625 milliseconds as
`being just one value that was found to be effective experimentally (“a good
`value”) in non-stationary noise settings. We, therefore, find unpersuasive
`Patent Owner’s position that 225 milliseconds is a significant difference
`when the record lacks evidence of how or if 225 milliseconds would impact
`Hirsch’s system’s performance in non-stationary noise settings (i.e., there is
`no experimental determination in Hirsch as to what would be effective in
`non-stationary settings and as Martin explained, these values are determined
`experimentally).
`But even if we accept Patent Owner’s position that Hirsch’s 225
`millisecond difference is significant, and assume (without any factual basis
`from Patent Owner) that such a difference would remain even in non-
`stationary noise settings, that is still not a reason to find that Hirsch teaches
`
`11
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`away from Martin. Based on the evidence before us, we have no way to
`compare Hirsch and Martin with respect to effectiveness in non-stationary
`noise settings. But between Hirsch and Martin, Martin is the only reference
`that expressly states its effectiveness in non-stationary noise settings. See
`Ex. 1006, 1096; PO Remand Br. 5 (Patent Owner acknowledging “[n]othing
`in Hirsch discusses how its algorithm might perform in non-stationary noise
`environments.”). That is, Hirsh’s disclosure of a shorter window length for
`stationary noise settings does not disparage or discourage the use of Martin’s
`techniques, known to be effective for non-stationary noise settings, even if
`Martin’s technique uses a window length that is 225 milliseconds longer
`than Hirsch’s.
`And, again, assuming that the shorter timeframe in Hirsch applies to
`non-stationary noise settings, it is not necessarily an improvement over
`Martin. Rather, one skilled in the art would have appreciated the benefits of
`a proven system to handle an identified problem (e.g., the algorithm of
`Martin addressing the problem identified in Hirsch) as a tradeoff to the
`unproven system of Hirsch in non-stationary noise settings (at least based on
`the evidence before us) with potentially shorter sample times required. See
`Allied Erecting and Dismantling Co. v. Genesis Attachments, LLC, 825 F.3d
`1373, 1381 (Fed. Cir. 2016) (“A given course of action often has
`simultaneous advantages and disadvantages, and this does not necessarily
`obviate motivation to combine.”) (quoting Medichem, S.A. v. Rolabo, S.L.,
`437 F.3d 1157, 1165 (Fed. Cir. 2006)).
`As Petitioner notes, “Dr. Hochwald [testifies that] Hirsch’s prior
`article [(Ex. 1032)] recognizes that there is a tradeoff between the accuracy
`of a noise estimate and the ability to adapt to the noise and that the optimum
`choice of window depends on the nature of the speech signals being
`
`12
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`considered.” Pet. Reply 17 (citing Ex. 1023 ¶¶ 33, 37). We credit the
`testimony of Dr. Hochwald because it is consistent with and supported by
`the disclosure of the prior article referenced in that testimony. See Ex. 1032,
`24 (discussing window length relative to accuracy).
`The preponderance of the evidence supports Petitioner’s position that
`Martin provides an effective noise floor algorithm in non-stationary noise
`settings and that Hirsch would not discourage one skilled in the art from
`looking to Martin’s teachings regarding that algorithm. Rather, based on the
`express reference to Martin in Hirsch, Martin’s discussion of its favorable
`results in non-stationary noise situations, and Hirsch’s silence with respect to
`its effectiveness in non-stationary noise settings, one skilled in the art would
`have been motivated to look to Martin’s noise floor algorithm.
`3. Whether Hirsch obviates the need to address the speech pause detection
`problem and, therefore, a skilled artisan would not have looked to
`Martin to address the problem
`Patent Owner additionally contends that “a skilled artisan would not
`look to this disclosure in Martin as reason to improve Hirsch because
`Hirsch’s techniques similarly obviate the need for speech pause detection,”
`but acknowledges that “Hirsch notes that Martin’s algorithm avoids the
`problem of speech pause detection.” PO Resp. 36. For the various reasons
`explained above, we do not agree that Hirsch obviates the need to look to
`Martin’s algorithm.
`As explained above, Hirsch notes the difficulties typically associated
`with non-stationary background noise, but does not specifically discuss the
`effectiveness of its algorithm in non-stationary noise settings, which is the
`purpose for Petitioner’s proposed combination with Martin’s teachings.
`Petitioner contends that “[t]he practices followed by the Hirsch authors is
`
`13
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`consistent [with] the general approach followed by those working in the field
`of noise reduction, where techniques addressing particular challenges or
`issues (e.g., noise reduction in nonstationary sound environments) are
`routinely added to or used to modify an existing noise reduction scheme.”
`Pet. 35 (citing Ex. 1003 ¶¶ 125–126). Petitioner’s contentions regarding
`“the general approach followed by those working in the field of noise
`reduction” noted above are supported by the testimony of Dr. Hochwald
`noted above (Ex. 1003 ¶¶ 125–26) and are not disputed by Patent Owner.
`Because Martin provides an effective approach for non-stationary
`noise settings, and similarly avoids the need for speech pause detection, we
`agree with Petitioner’s contentions that one skilled in the art would have
`combined Martin’s teachings with those of Hirsch, at least for non-stationary
`noise settings. Hirsch identifies problems with speech pause detection, but
`discusses effectiveness of its system only with respect to stationary noise
`settings. We do not agree that this obviates the need to address speech pause
`detection in all settings, particularly when Hirsch, itself, identifies the
`difficulties associated with non-stationary noise settings, but fails to address
`effectiveness of its system in such settings. See Ex. 1005, 153–154.
`
`14
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`4. How the teachings of Hirsch and Martin are combined
`The manner in which the teachings of Hirsch and Martin are
`combined (i.e., how the teachings are combined) does not appear to be
`within the scope of this remand. As explained above, in its decision, the
`Federal Circuit understood the “how” to be “specifically incorporat[ing]
`Martin’s noise power estimation algorithm into Hirsch’s spectral subtraction
`system or threshold calculation” and remanded for us to provide specific
`findings related to why a one skilled in the art would have made that
`modification. See Andrea, 2022 WL 1197341, at *5.
`Nevertheless, to the extent we misunderstood the Federal Circuit’s
`remand instructions, and “how” the teachings of Hirsch and Martin are
`combined is still at issue, we address Patent Owner’s contention that
`“[Petitioner] fails to sufficiently explain how or why various elements from
`the two systems described in Hirsch and Martin would be combined in the
`manner as the elements are arranged in the claim.” PO Resp. 40. Patent
`Owner refers to various portions of Dr. Hochwald’s testimony, but does not
`allege any specific fault in the Petition. See id. at 40–43.
`As noted above, Petitioner explains that the proposed modification
`“would have been to incorporate Martin’s technique for estimating the noise
`floor of a signal within Hirsch’s adaptive threshold calculation to improve
`operation of Hirsch’s system in non-stationary noise settings.” Pet. 35; see
`also Pet. Reply 20 (“The one simple alteration, as Dr. Hochwald explained,
`is modifying Hirsch’s adaptive threshold calculation ‘to use Martin’s noise
`floor algorithm instead of Hirsch’s running average algorithm for deciding
`the threshold in each frequency bin’” with “[t]he other steps in Hirsch . . .
`remain[ing] unaltered.”) (citing Ex. 1003 ¶¶ 127–130). We determine that
`this explanation by Petitioner sufficiently explains how to combine the
`
`15
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`teachings of Hirsch with those of Martin. And as explained above, the
`Federal Circuit understood this to be the “how” as well. See Andrea, 2022
`WL 1197341, at *5.
`According to Petitioner, and we agree, “[a] skilled person would have
`recognized the Martin technique could be used in place of the analogous
`determination being made in Hirsch for each frequency bin.” Pet. 37 (citing
`Ex. 1003 ¶ 130). We understand Petitioner’s proposed modification to be a
`substitution specifically of Martin’s noise floor algorithm into Hirsch’s
`system. As noted above, Petitioner’s proposed modification is simply “the
`general approach followed by those working in the field of noise reduction,
`where techniques addressing particular challenges or issues (e.g., noise
`reduction in nonstationary sound environments) are routinely added to or
`used to modify an existing noise reduction scheme.” Pet. 35 (citing
`Ex. 1003 ¶¶ 125–126). That is, this proposed modification would have been
`within the level of skill in the art and would have been met with a reasonable
`expectation of success. See Amgen Inc. v. F. Hoffman-LA Roche Ltd., 580
`F.3d 1340, 1362 (Fed. Cir. 2009) (“An obviousness determination requires
`that a skilled artisan would have perceived a reasonable expectation of
`success in making the invention in light of the prior art.”). Patent Owner
`does not dispute Petitioner’s characterization of the common practices in the
`art.
`
`Based on the record before us, Petitioner provides sufficient
`explanation of how the alleged teachings of Hirsch and Martin would have
`been combined in the proposed modification as noted above.
`D. Summary
`As explained above, Hirsch notes the difficulty of addressing non-
`stationary noise environments, but does not provide results regarding the
`
`16
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`performance of its system in such environments. Martin provides a noise
`floor algorithm that provides good results in non-stationary noise
`environments. Based on the record before us, one skilled in the art would
`have appreciated that Martin provides a proven way to deal with non-
`stationary noise environments that one skilled in the art would have been
`motivated to substitute into Hirsch’s system based on Hirsch’s identification
`of this known problem and identification of Martin as a known solution to
`this problem. Consequently, Petitioner has presented sufficient evidence to
`support its reasoning as to why one skilled in the art would have been
`motivated to combine the teachings of Hirsch and Martin with respect to
`claims 6–9.
`
`III. CONCLUSION3
`
`In summary:
`Claims
`35
`
`U.S.C. §
`
`Reference(s)/Basis
`
`103
`
`Hirsch, Martin
`
`6–9
`
`
`
`
`Claims
`Shown
`Unpatentable
`6–9
`
`Claims
`Not shown
`Unpatentable
`
`
`IV. ORDER
`In consideration of the foregoing, it is hereby:
`
`
`3 Should Patent Owner wish to pursue amendment of the challenged claims
`in a reissue or reexamination proceeding subsequent to the issuance of this
`Decision, we draw Patent Owner’s attention to the April 2019 Notice
`Regarding Options for Amendments by Patent Owner Through Reissue or
`Reexamination During a Pending AIA Trial Proceeding. See 84 Fed. Reg.
`16,654 (Apr. 22, 2019). If Patent Owner chooses to file a reissue application
`or a request for reexamination of the challenged patent, we remind Patent
`Owner of its continuing obligation to notify the Board of any such related
`matters in updated mandatory notices. See 37 C.F.R. § 42.8(a)(3), (b)(2).
`
`17
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`ORDERED that Petitioner has established by a preponderance of the
`evidence that claims 6–9 of the ’345 patent are unpatentable; and
`FURTHER ORDERED that, because this is a Final Written Decision
`on Remand, parties to this proceeding seeking judicial review of the
`Decision must comply with the notice and service requirements of 37 C.F.R.
`§ 90.2.
`
`
`
`18
`
`
`
`IPR2017-00626
`Patent 6,363,345 B1
`FOR PETITIONER:
`
`Jeffrey Kushan
`Steven Baik
`Thomas Broughan
`SIDLEY AUSTIN LLP
`jkushan@sidley.com
`sbaik@sidley.com
`tbroughan@sidley.com
`
`
`FOR PATENT OWNER:
`
`William Belanger
`Andrew Zappia
`TROUTMAN PEPPER
`william.belanger@troutman.com
`andrew.zappia@troutman.com
`
`
`Andrew Schultz
`Griffin Mesmer
`Frank Liu
`PEPPER HAMILTON LLP
`schultza@pepperlaw.com
`mesmerg@pepperlaw.com
`liuf@pepperlaw.com
`
`
`
`19
`
`