Paper No. 18
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`____________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`____________________
`
`
`APPLE INC.
`Petitioner,
`
`v.
`
`ANDREA ELECTRONICS INC.,
`Patent Owner.
`
`Patent No. 6,363,345
`____________________
`
`Inter Partes Review No. IPR2017-00627
`__________________________________________________________________
`
`Petitioner’s Reply
`
`
`
`
`
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Table of Contents
`
`2.
`
`Introduction .................................................................................................... 1
`I.
`II. Claim Construction ....................................................................................... 2
`III. Helf Anticipates Claims 1-3, 13-14, 21, 23, and 38 ..................................... 3
`IV. Helf Alone Anticipates or Renders Obvious Claims 4-7, 9-11, 39-41, and
`43 ..................................................................................................................... 3
`A. Helf Alone Discloses or Renders Obvious the Elements of Claims
`4 and 39 ................................................................................................. 4
`1.
`Helf Discloses or Renders Obvious the “Future Minimum” ...... 5
`2.
`Helf Discloses or Renders Obvious “Deriving” the Current
`Minimum in Accordance with a “Future Minimum” ................. 8
`B. Helf Alone Discloses or Renders Obvious the Elements of Claims
`6, 7, and 9 .............................................................................................. 9
`1.
`Helf Discloses or Renders Obvious Updating the Future
`Minimum “Periodically” as Required by Claim 6 ...................... 9
`Helf Discloses or Renders Obvious the “Current Magnitude”
`and Other Elements of Claims 7 and 9 ..................................... 10
`C. Helf Alone Discloses or Renders Obvious Claims 4-7, 9-11, 39-41,
`and 43 .................................................................................................. 11
`V. Helf and Martin Render Claims 6, 8, 9, 12, 25, 42, and 46 Obvious ...... 11
`A. A POSA Would Have Considered Helf Together with Martin .... 11
`B. Helf and Martin Render Claims 6, 8, and 9 Obvious .................... 15
`1.
`Helf and Martin Render Obvious Setting the Current Minimum
`to the Future Minimum “Periodically” per Claim 6 ................. 16
`Helf and Martin Render Obvious Setting the Current Minimum
`Equal to the Current Magnitude per Claim 8 ............................ 17
`Helf and Martin Render Obvious Updating the Future
`Minimum “Periodically” per Claim 9 ....................................... 19
`VI. A POSA Would Have Considered it Obvious to Modify Helf Alone or
`with Martin with Conventional Spectral Subtraction Techniques ......... 20
`
`2.
`
`3.
`
`
`
`ii
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`A. A POSA Would Have Combined Helf with Boll ............................ 20
`B.
`A POSA Would Have Combined Helf with Arslan and Boll ........ 22
`C. A POSA Would Have Combined Helf with Martin and Uesugi ... 24
`VII. Conclusion .................................................................................................... 25
`
`
`
`
`
`iii
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`TABLE OF AUTHORITIES
`
`Page(s)
`
`Cases
`In re Keller,
`642 F.2d 413 (CCPA 1981) .......................................................................... 14, 22
`In re Nievelt,
`482 F.2d 965 (CCPA 1973) ................................................................................ 22
`In re Sneed,
`710 F.2d 1544 (Fed. Cir. 1983) .......................................................................... 22
`
`
`
`
`
`
`iv
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Exhibit List
`
`Reference Name
`U.S. Patent No. 6,363,345
`U.S. Patent No. 6,363,345 File History
`Declaration of Bertrand Hochwald
`[Reserved]
`H. G. Hirsch and C. Ehricher, “Noise estimation techniques for
`robust speech recognition,” Proc. IEEE Int. Conf. Acoustics,
`Speech, Signal Processing, vol. 1, pp. 153 -156, 1995
`(“Hirsch”)
`Rainer Martin, “An Efficient Algorithm to Estimate the
`Instantaneous SNR of Speech Signals,” Proc. Eurospeech, pp.
`1093-96, 1993 (“Martin”)
`Letter from Technische Informationsbibliothek re: Proc.
`Eurospeech 1993 (2 Jan. 2017)
`Proc. Eurospeech 1993 Vol. 2 Table of Contents from
`Technische Informationsbibliothek
`Steven F. Boll, “Suppression of Acoustic Noise in Speech
`Using Spectral Subtraction,” IEEE Transactions on Acoustics,
`Speech, and Signal Processing, Vol. ASSP-27, No. 2, April
`1979 (“Boll”)
`U.S. Patent No. 5,550,924 to Helf (“Helf”)
`U.S. Patent No. 5,706,395 to Arslan (“Arslan”)
`Excerpts from Deller et al., Discrete-Time Processing of Speech
`Signals (1993)
`Excerpt from Merriam-Webster Dictionary (1993)
`Excerpts from Oppenheim and Willsky, Signals and Systems
`(1997)
`U.S. Patent No. 5,459,683 to Uesugi
`Lim and Oppenheim, “Enhancement and Bandwidth
`Compression of Noisy Speech,” Proceedings of the IEEE, vol.
`67, no. 12, pp. 1586-1604, December 1979
`Affidavit of Service in Andrea Elecs. v. Apple Inc., EDNY
`
`Exhibit #
`1001
`1002
`1003
`1004
`1005
`
`1006
`
`1007
`
`1008
`
`1009
`
`1010
`1011
`1012
`
`1013
`1014
`
`1015
`1016
`
`1017
`
`
`
`v
`
`

`

`1019
`
`1020
`
`1021
`
`1022
`
`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Exhibit #
`1018
`
`Reference Name
`In the Matter of Certain Audio Processing Hardware and
`Software and Products Containing the Same, Inv. No. 337-TA-
`949, Claim Construction Order (U.S.I.T.C. Jan. 27, 2016) (“949
`CC Order”)
`In the Matter of Certain Audio Processing Hardware and
`Software and Products Containing Same, Inv. No. 337-TA-949,
`Complainant Andrea Electronics Corp.’s Initial Claim
`Construction Brief (U.S.I.T.C. Oct. 19, 2015) (“Andrea CC
`Br.”)
`In the Matter of Certain Audio Processing Hardware and
`Software and Products Containing Same, Inv. No. 337-TA-949,
`Commission Investigative Staff’s Initial Markman Brief
`(U.S.I.T.C. Oct. 19, 2015) (“OUII CC Br.”)
`Letter from the parties in 337-TA-949 informing ALJ they
`agreed to certain constructions (Nov. 10, 2015) (prior litigation)
`In the Matter of Certain Audio Processing Hardware, Software,
`and Products Containing The Same, Inv. No. 337-TA-1026,
`Verified Complaint Against Apple Inc. and Samsung Inc.
`Under Section 337 of the Tariff Act of 1930, as Amended
`(U.S.I.T.C. Sept. 19, 2016
`[NEW] 1023 Reserved
`[NEW] 1024 Reserved
`[NEW] 1025 Exhibit 2 from Hochwald Deposition
`[NEW] 1026 Transcript from Deposition of Scott Douglas dated Jan. 17,
`2018
`[NEW] 1027 Exhibit 1 from Douglas Dep., Figure 27 depicting Current and
`Future Minima
`[NEW] 1028 Exhibit 2 from Douglas Dep., Dr. Douglas’s mark up of Exhibit
`1
`[NEW] 1029 Exhibit 8 from Douglas Dep., Declaration of Scott Douglas in
`Support of Complainant Andrea’s Claim Construction Brief in
`Inv. No. 337-TA-949 (Oct. 19, 2015)
`[NEW] 1030 Transcript from Deposition of Scott Douglas dated June 16,
`2017, taken in Inv. No. 337-TA-1026
`
`
`
`vi
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Reference Name
`Exhibit #
`[NEW] 1031 Rainer Martin, Spectral Subtraction Based on Minimum
`Statistics, Proc. EUSIPCO-94, pp. 1182-85 (1994) (“Martin
`94”)
`[NEW] 1032 H. G. Hirsch, “Estimation of Noise Spectrum and its
`Application to SNR Estimation and Speech Enhancement,”
`Technical Report TR-93-012, International Computer Science
`Institute (1993) (reference [7] in Martin 93)
`[NEW] 1033 D. Van Campernolle, “Noise Adaptation in a Hidden Markov
`Model Speech Recognition System”, Computer Speech and
`Language, Vol. 3, pp. 151-167 (1989) (reference [3] in Martin
`93)
`
`vii
`
`
`
`
`
`
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`I.
`
`Introduction
`Patent Owner Andrea raises no challenge to the Board’s initial finding that
`
`Helf anticipates claims 1-3, 13-14, 21, 23, and 38, effectively conceding that these
`
`claims are unpatentable. See Paper 11 (“Resp.”), 12. Andrea instead devotes the
`
`bulk of its Response to arguing that dependent claims 4-11 and 39-41—which
`
`cover a process for tracking the noise floor of an audio signal—are patentable over
`
`Helf alone or in combination with Martin. See Resp., 17-51. Andrea’s primary
`
`argument is that Helf does not teach a “future minimum” value because Helf selects
`
`an average value as the minimum of a frequency bin, and an “average” allegedly
`
`cannot be a “minimum.” But the ’345 specification discloses the exact same
`
`operation: setting the “future minimum” to an average magnitude. Ex. 1001, 6:25-
`
`26 (“A future minimum value is initiated… with the value of the current
`
`magnitude (Y(n))”), 7:28-30 (“Note also that the values of Y in the figures are the
`
`smoothed values of Y after averaging.”). Like the rest of Andrea’s arguments, its
`
`primary purported distinction between Helf and the ’345 claims lacks any merit.
`
`Notably, Andrea does not dispute that the other claims (which all depend
`
`from claims 1 or 38) simply recite conventional features of the spectral subtraction
`
`process or that Helf and the secondary references disclose these limitations.
`
`Instead, Andrea contends only that a person of ordinary skill in the art (“POSA”)
`
`would not have been motivated to modify Helf to incorporate these known and
`
`
`
`1
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`conventional features. See Resp., 51-59. For the reasons Apple provided in its
`
`Petition, a POSA would have had ample reason to modify Helf to use these
`
`conventional features to solve standard problems present in any spectral
`
`subtraction system. Accordingly, the Board should cancel claims 1-25 and 38-47.
`
`II. Claim Construction
`The Board need not adopt specific constructions for most terms because
`
`under any reasonable construction, Helf either anticipates or renders the claims
`
`obvious.
`
`In its Response, Andrea proposes construing a single term, contending that
`
`“periodically” in claims 6 and 9 should mean “at regular intervals,” thereby
`
`limiting the term to mean the first of two dictionary definitions that Apple
`
`presented. See Resp., 6-9. If the Board determines that it needs to construe
`
`“periodically,” that term’s broadest reasonable construction also includes the
`
`second dictionary definition, meaning “from time to time”.1 See Ex. 1013.
`
`Andrea’s argument for restricting the scope of “periodically” to mean only “at
`
`regular intervals” improperly imports limitations from an example in the patent
`
`
`
`1 As explained below, Helf discloses the “periodically” limitation under the term’s
`
`broader meaning, and it renders that limitation obvious under Andrea’s narrower
`
`proposed interpretation.
`
`
`
`2
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`specification. The ’345 specification states that the minimum values are
`
`“calculated continuously and updated periodically, for example, every 5 seconds.”
`
`Ex. 1001, 8:36-40; see id., 3:32-45, 8:64-67 (“while specific values are used [] in
`
`the several equations and calculations employed in the present invention, these
`
`values may be different than those shown”). The specification expressly states that
`
`period is exemplary, and nothing indicates that the update period must always be
`
`the same or that it cannot change as the noise reduction process runs.
`
`III. Helf Anticipates Claims 1-3, 13-14, 21, 23, and 38
`Andrea does not dispute that Helf anticipates claims 1-3, 13-14, 21, 23, and
`
`38, which includes both challenged independent claims (1 and 38). Resp., 10. The
`
`Board should therefore cancel these claims for the reasons set forth in the Petition.
`
`See Pet., 27-32.
`
`IV. Helf Alone Anticipates or Renders Obvious Claims 4-7, 9-11, 39-41, and
`43
`Andrea raises several challenges to claims 4-7, 9-11, 39-41, and 43, focusing
`
`on limitations in four specific claims: 4 (and parallel method claim 39), 6, 7, and
`
`9.2 Andrea’s arguments lack merit.
`
`
`
`2 Andrea does not advance any unique arguments for claims 5, 7, 10-11, 25, 40-42,
`
`or 46, and consequently, Apple does not specifically address those claims here.
`
`
`
`3
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`A. Helf Alone Discloses or Renders Obvious the Elements of Claims
`4 and 393
`The Board correctly found that Helf anticipates or renders obvious claims 4
`
`and 39. Claim 4 depends from claim 1 and specifies “set[ting] the threshold… in
`
`accordance with a current minimum value… of the corresponding frequency bin;
`
`said current minimum value being derived in accordance with a future minimum
`
`value… of the corresponding frequency bin.” Claim 39 depends from claim 38 and
`
`specifies the same limitation.
`
`In the Petition, Apple explained how Helf mapped to the elements of claims
`
`4 and 39. Pet., 32-34; Ex. 1010(Helf), 8:10-59. Helf discloses a running minimum
`
`noise estimator that tracks a noise estimate Mk as the minimum value of a
`
`frequency bin over a 10 second period. Pet., 33-34; Ex. 1010(Helf), 8:10-32. The
`
`estimate Mk corresponds to the “future minimum” value. Pet., 33. Helf tracks a
`
`noise estimate Nk (also denoted interchangeably as Ck) that is used to set several
`
`noise thresholds, and Nk corresponds to the “current minimum.” Pet., 32-33; Ex.
`
`1010(Helf), 9:16-25, 9:39-65. Helf will set the noise estimate Nk equal to the
`
`
`
`3 A heading in Andrea’s Response asserts that Martin does not disclose either the
`
`“current minimum” or “future minimum” element, but the Response only
`
`substantively addresses the “future minimum” element.
`
`
`
`4
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`minimum Mk when the sum of the differences between Nk and Mk across all
`
`frequency bins exceeds a threshold D, and thus “derives” the “current minimum”
`
`Nk in accordance with the “future minimum” Mk as required by claims 4 and 39.
`
`Pet., 33-34; Ex. 1010(Helf), 8:38-59.
`
`In its Response, Andrea contends that Helf does not disclose a “future
`
`minimum” value and that Helf does not show “deriving” a current minimum value
`
`in accordance with a future minimum value. Neither argument has merit.
`
`1. Helf Discloses or Renders Obvious the “Future Minimum”
`Andrea contends that Helf does not disclose the “future minimum” value,
`
`raising several arguments. First, Andrea asserts that the noise estimate Bk that is
`
`generated by Helf’s stationary noise estimator is not a minimum value, and
`
`therefore cannot be the “future minimum.” Resp., 18-19. Apple does not rely on
`
`Bk as the “future minimum” so Andrea’s argument regarding Bk is irrelevant.
`
`Instead, Apple relies on Mk which is generated by Helf’s running minimum
`
`estimator. Pet., 32-33. As Helf explains, in some scenarios, all noise estimates
`
`will be provided by the running minimum estimator: “[t]here will be some
`
`instances when… the stationary noise estimator… will never produce noise
`
`background estimates. For these cases, the running minimum estimator 22 will
`
`produce noise background estimates.” Ex. 1010(Helf), 8:11-17. Thus, there are
`
`scenarios where the stationary estimator and Bk are never used.
`
`
`
`5
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Next, Andrea argues that minimum estimate Mk is not a “future minimum”
`
`because Mk is calculated as the average value of a frequency bin over 8 frames,
`
`and an average of a set of numbers will almost always be greater than the
`
`minimum value of that set of numbers. Resp., 19-21. Andrea’s assertion that an
`
`average cannot be a minimum is inconsistent with the claim language and the ’345
`
`specification.
`
`The ’345 claims expressly contemplate that an average value can be a
`
`minimum value. For example, claim 39 specifies tracking a “future minimum
`
`value of the magnitude of [a] frequency bin,” and claim 45 (which depends
`
`through claim 39) specifies that the magnitude values of each frequency bin can be
`
`estimated and smoothed (i.e., averaged). Ex. 1001, 12:27-29, 12:50-54 (claim 44
`
`specifying “the step of estimating a magnitude of each frequency bin” and claim 45
`
`specifying “the step of smoothing the estimate of each frequency bin”). The ’345
`
`specification shows that “smoothing” means averaging. Ex. 1001, 7:28-30.
`
`The ’345 specification also shows embodiments where the “future
`
`minimum” is set as an average value. E.g., Ex. 1001, 7:28-30. For example, figure
`
`7 shows the frequency bin magnitude Y(n) is calculated as an average across
`
`frequency bins (box 706) and over time (box 708) and that average value is used to
`
`set the future minimum value (box 716).
`
`
`
`6
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`
`
`The specification explains that each bin’s current magnitude is denoted
`
`“Y(n),” (Ex. 1001, 6:25-26 (“A future minimum value is initiated… with the value
`
`of the current magnitude (Y(n))”), 6:46, 7:9), and it shows that the Y(n) values
`
`can be calculated as an average, Ex. 1001, 7:28-30 (“Note also that the values of Y
`
`in the figures are the smoothed values of Y after averaging”), Figs. 2 (calculating
`
`Y(n) as an average), 3 (setting future minimum to Y(n) values from Fig. 2), 7 (all
`
`emphases added).
`
`Helf works the same way as the ’345 patent. During each frame, Helf
`
`calculates the average signal power over the past 8 frames (denoted Mk(fk)). Ex.
`
`1010(Helf), 8:19-32; Pet., 33; compare with Ex. 1001, Fig. 7 (boxes 706 and 708).
`
`
`
`7
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Helf then sets Mk equal to the minimum of all the average values Mk(fk). Ex.
`
`1010(Helf), 8:19-32. That the value Mk is the minimum average value is irrelevant
`
`to the claims.
`
`Andrea argues that the averaging done by Helf is different than the
`
`averaging done in the ’345 patent because Helf allegedly calculates just one
`
`average value. Resp., 20-23. But Helf calculates the average magnitude for each
`
`frame fk, and it selects the minimum estimate Mk as the smallest of all Mk(fk)
`
`values. Ex. 1010(Helf), 8:19-32; Ex. 1004, ¶111. Thus, Helf works the same way
`
`as the ’345 patent, and Andrea’s attempt to differentiate the averaging done in the
`
`’345 patent from that done in Helf fails.
`
`2. Helf Discloses or Renders Obvious “Deriving” the Current
`Minimum in Accordance with a “Future Minimum”
`Andrea argues Helf’s noise estimate Nk (“current minimum”) is not
`
`“derived” in accordance with Mk (“future minimum”) because Helf sets Nk equal to
`
`Mk only when the sum of the differences between Mk and Nk across every
`
`frequency bin in the frame is less than a threshold D. Resp., 23-24. According to
`
`Andrea, this means that Nk is derived in accordance with the Mk value of every
`
`frequency bin and not the “future minimum of the corresponding frequency bin.”
`
`Resp., 23-24. Andrea is incorrect. For each frequency bin, Helf’s Nk is “derived”
`
`from the Mk value for that bin because Nk is set equal to Mk. Ex. 1004, ¶¶130-32.
`
`That this update is conditioned upon the outcome of a test is irrelevant to the claim
`
`
`
`8
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`language, because Andrea has not shown that the broadest reasonable construction
`
`of the claim language “derived in accordance with” must be limited to “set equal
`
`to without any other calculations, or consideration of other values.” In their
`
`broadest reasonable construction, the claims encompass Helf’s process of setting
`
`Nk set equal to Mk after considering the threshold D.
`
`B. Helf Alone Discloses or Renders Obvious the Elements of Claims
`6, 7, and 9
`1. Helf Discloses or Renders Obvious Updating the Future
`Minimum “Periodically” as Required by Claim 6
`Andrea incorrectly asserts that Helf does not disclose that the “current
`
`minimum” Nk is set to the “future minimum” Mk “periodically” as required by claim
`
`6 because Helf’s running minimum estimator is not always invoked. Resp., 25.
`
`Andrea’s argument ignores Helf’s teaching that, in some scenarios, all noise
`
`estimates will be provided by the running minimum estimator. Ex. 1010(Helf),
`
`8:11-17. In those scenarios, Helf always sets the noise estimate Nk equal to Mk 10
`
`seconds after the algorithm starts running. Ex. 1004, ¶¶109, 128, 133. After that
`
`happens, Helf then will set Nk equal to Mk whenever the sum of the differences
`
`between Mk and Nk across every frequency bin in the frame is greater than a
`
`threshold D, which happens from time to time or “periodically.” Ex. 1004, ¶¶130-
`
`
`
`9
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`32. Under the broadest reasonable interpretation of “periodically,” Helf anticipates
`
`claim 6.4
`
`2. Helf Discloses or Renders Obvious the “Current
`Magnitude” and Other Elements of Claims 7 and 9
`For claims 7 and 9, Andrea asserts that Helf does not set its future minimum
`
`Mk equal to a current magnitude value because Mk is set to the minimum average
`
`value. Resp., 27. But as explained with the “future minimum” in claim 4
`
`(§IV.A.1, above), nothing in the claims precludes setting the “future minimum” to
`
`a “current magnitude” value that is calculated as an average value. The
`
`specification explains that each bin’s current magnitude is denoted “Y(n),” (Ex.
`
`1001, 6:25-26, 6:46, 7:9), and it shows that the Y(n) values can be calculated as an
`
`average, Ex. 1001, 7:28-30, Figs. 2 (calculating Y(n) as an average), 3 (setting
`
`future minimum to Y(n) values from Fig. 2), 7 (setting future minimum to Y(n)
`
`which is calculated as an average).
`
`For claim 9, Andrea asserts that Helf does not disclose setting the future
`
`minimum Mk to the current magnitude “periodically” relying on a narrow
`
`interpretation of “periodically” to mean at regular intervals. Helf shows that
`
`
`
`4 If “periodically” is interpreted more narrowly, claim 6 would have been obvious
`
`as explained below.
`
`
`
`10
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`whenever the minimum average magnitude Mk(fk) (“current magnitude”) of the
`
`current frame fk is less than the future minimum Mk, it updates Mk to be the
`
`minimum average magnitude. Ex. 1004, ¶¶111, 134. Thus, Helf shows updating
`
`the future minimum Mk from time to time, which satisfies the broadest reasonable
`
`interpretation of “periodically.”5
`
`C. Helf Alone Discloses or Renders Obvious Claims 4-7, 9-11, 39-41,
`and 43
`As explained in the Petition, Helf shows performing certain noise processing
`
`operations on signal power instead of signal magnitude, but that modifying Helf to
`
`perform those operations using signal magnitude would have been obvious. Pet.,
`
`38-41. Patent Owner explicitly states that it does not challenge that Helf discloses
`
`or renders obvious performing its operations on signal magnitude. Resp., 10 n.1.
`
`V. Helf and Martin Render Claims 6, 8, 9, 12, 25, 42, and 46 Obvious
`A. A POSA Would Have Considered Helf Together with Martin
`Andrea contests that a POSA would have combined the teachings of Helf
`
`and Martin based on several factually incorrect arguments. First, Andrea alleges a
`
`POSA would not have considered Helf and Martin together because of
`
`“fundamental” differences between the two. Resp., 49; id., 46, 48-50. Contrary to
`
`
`
`5 If “periodically” is given Andrea’s narrow interpretation, claim 9 would have
`
`been obvious as explained below.
`
`
`
`11
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Andrea’s assertions, no such fundamental differences exist. For example, Andrea
`
`asserts Helf is “fundamentally” different from Martin because Helf calculates the
`
`minimum value for a frequency bin as the minimum of an average signal value and
`
`not the signal value itself. Resp., 49; see Ex. 1010(Helf), 8:19-31 (calculating each
`
`Mk(fk) as average over 8 frames). But Martin does the same thing. Martin smooths
`
`or averages the samples over time (e.g., over 128 frames), and then tracks the
`
`minimum smoothed power estimate. Ex. 1006(Martin), 1094 (“The minimum
`
`power of the last M samples is found by a samplewise comparison of the actual
`
`minimum PMmin(i) and the smoothed power (cid:2172)(cid:3365)x(i).” (emphasis added)). And both
`
`Helf and Martin use the smoothed values to determine a minimum value which is
`
`used to calculate a noise estimate. Ex. 1010(Helf), 8:33-36 (“Use the minimum
`
`values Mk… as the background noise spectral estimate if… two conditions are both
`
`met”); Ex. 1006(Martin), 1093 (“To estimate the noise floor our algorithm takes
`
`the minimum of a smoothed power estimate within a window”). Thus, the use of
`
`average values is a commonality, not a difference, between Helf and Martin.
`
`Andrea also asserts Martin’s calculation of noise statistics differentiates it
`
`from Helf, Resp., 48, but Andrea is incorrect as Helf also estimates noise
`
`statistics. Ex. 1010(Helf), 9:32-46 (local speech versus noise detector calculates a
`
`probability or “confidence” that a frequency bin contains noise). In addition,
`
`Andrea’s current description of Helf as not involving statistics is directly contrary
`
`
`
`12
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`to the position it took in its Preliminary Response, where it admitted Helf
`
`estimates noise “probability” (i.e., statistics). Paper 6 at 30 (“equation (12) of
`
`Helf is used to determine the probability or likelihood that a frequency component
`
`contains noise” (emphasis added)).
`
`Andrea also asserts Martin differs from Helf because Martin uses signal-to-
`
`noise ratio (SNR). Resp., 48-49. But Andrea ignores that Helf also uses signal-to-
`
`noise ratio. Ex. 1010(Helf), 9:39-42 (“decision are based on the ratio of the
`
`magnitude of the [signal] to the magnitude of the… background noise”), 9:47-57
`
`(showing calculations based on the ratio of signal (Sk) to noise (Nk), “Sk/Nk”).
`
`Thus, each of the differences Andrea identifies actually is a commonality between
`
`Helf and Martin.
`
`Next, Andrea argues that the skilled person would not have been motivated
`
`to combine Helf and Martin. Resp., 46. Andrea is wrong. As Dr. Hochwald
`
`explained:
`
`As part of the normal development process, a person of ordinary skill
`would have looked for ways to customize or optimize the performance
`of a technique to fit the particular details of a particular system. In
`doing so, a person of ordinary skill would have considered features
`and techniques described in related references.
`
`Ex. 1004, ¶153. Thus, the skilled person would have recognized that the
`
`parameters of Helf’s noise processing algorithm could be adjusted, and that person
`
`
`
`13
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`would have been motivated to look to a related reference like Martin for guidance
`
`about what parameters may be beneficial. Ex. 1004, ¶¶154-55. Dr. Hochwald
`
`further explained that the skilled person considering Helf and Martin would have
`
`recognized it was beneficial to try making certain adjustments to Helf based on
`
`Martin, and identified several such modifications. Ex. 1004, ¶¶159, 162; Pet., 44-
`
`52.
`
`Next, Andrea argues that Apple did not specify what aspects of Martin
`
`would be incorporated into Helf. Resp., 48. But Apple did not argue that Martin’s
`
`algorithm could be bodily incorporated into Helf, and thus, no specific portions of
`
`Martin’s algorithm are literally inserted into Helf. In re Keller, 642 F.2d 413, 425
`
`(CCPA 1981) (“The test for obviousness is not whether the features of a secondary
`
`reference may be bodily incorporated into the structure of the primary reference....
`
`Rather, the test is what the combined teachings of those references would have
`
`suggested to those of ordinary skill in the art.”). Instead, Apple and Dr. Hochwald
`
`explained the skilled person would have adjusted some of the timing aspects of
`
`Helf’s algorithm based on features of Martin as this was part of the ordinary
`
`development process that would have been undertaken by a POSA. Pet.,43-44; Ex.
`
`1004, ¶¶155, 159-65. And with respect to the single hardware element—Martin’s
`
`microphone array—that Apple asserted would have been added to Helf, that
`
`element was clearly identified it as part of the combination. Pet., 51-52.
`
`
`
`14
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Andrea also alleges that that Martin’s techniques are not easily combinable
`
`with other systems. Resp., 48. Again, Apple did not argue bodily incorporation of
`
`Martin’s algorithm into Helf. Moreover, Andrea’s argument is premised on its
`
`assertion that “fundamental” differences exist between Helf and Martin, which is
`
`incorrect as explained above.
`
`Andrea argues a POSA would not be motivated to improve Helf’s running
`
`minimum estimator because Helf discloses that its stationary estimator is more
`
`accurate. Resp., 47, 49-50. But Helf discloses that during periods of continuous
`
`speech, only the running minimum estimator will be used. Ex. 1010(Helf), 8:11-
`
`17; Ex. 1004, ¶¶109-10, 128. Thus, as Dr. Hochwald explained, the POSA would
`
`have been motivated to improve that algorithm so that Helf’s noise estimation
`
`process was more accurate during such time of continuous speech. Ex. 1004, ¶162.
`
`B. Helf and Martin Render Claims 6, 8, and 9 Obvious
`While Andrea asserts the combination of Helf and Martin does not render
`
`any of claims 6, 8-9, 12, 25, and 46 obvious, it raises specific challenges only to
`
`claims 6, 8, and 9. Each of Andrea’s arguments rests on a misreading of the claim
`
`language or a mischaracterization of Helf or Martin and should be rejected.
`
`
`
`15
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`1. Helf and Martin Render Obvious Setting the Current
`Minimum to the Future Minimum “Periodically” per Claim
`6
`Apple explained that the skilled person would have modified Helf to
`
`periodically update the noise estimate Nk (“current minimum”) with the minimum
`
`estimate Mk (“future minimum”) every 10 seconds (“periodically”) based on
`
`Martin’s teaching of updating the noise estimate after every sub-window of M
`
`samples. Pet., 45-46. Andrea disagrees, raising two separate arguments.
`
`First, Andrea asserts that it would not have been obvious to modify Helf
`
`because Martin allegedly does not show updating a noise estimate “periodically”
`
`due to the fact that Martin shows that its noise estimate can sometimes be updated
`
`immediately. Resp., 39-40. Andrea ignores that Martin shows that the noise
`
`estimate Pn(i) is always updated at the end of every sub-window of M samples, and
`
`thus, it updates the noise estimate at regular intervals. Ex. 1004, ¶¶148, 164.
`
`There is no disagreement that Pn(i) is updated after every M samples, as Dr.
`
`Douglas admitted. Ex. 1026, 77:16-80:19. Though Martin does teach that the
`
`noise estimate can also be updated between these intervals, that does not change
`
`that Martin teaches that the noise estimate is always updated at the end of a sub-
`
`window.
`
`Second, Andrea argues that a POSA would not have tried to improve Helf’s
`
`running minimum estimator because Helf describes its stationary noise estimator
`
`
`
`16
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`as more accurate. Resp., 40-41. But Helf discloses that the stationary noise
`
`estimator cannot always be used, and consequently during certain periods, e.g.,
`
`during continuous speech, only the running minimum estimator can be used. Ex.
`
`1010(Helf), 8:11-17; Ex. 1004, ¶¶128, 162. Thus, the skilled person would have
`
`been motivated to improve the running minimum algorithm to increase the
`
`accuracy of Helf’s noise estimation process during such periods. Ex. 1004, ¶162.
`
`Andrea also argues that Apple’s modification to Helf’s running minimum
`
`estimator “abandons” Helf’s use of the threshold D to condition update of the noise
`
`estimate, and that a POSA would not have been motivated to make that
`
`modification. Resp., 41. But Apple’s modification does not require abandoning
`
`the use of the threshold D. Instead the modification supplements Helf’s use of D
`
`to also periodically (e.g., every 10 seconds) reset the noise estimate Nk to the
`
`minimum estimate Mk as taught by Martin. Pet., 46; Ex. 1004, ¶159. Andrea’s
`
`argument, thus, is irrelevant.
`
`2. Helf and Martin Render Obvious Setting the Current
`Minimum Equal to the Current Magnitude per Claim 8
`Apple explained that a POSA would have modified Helf to update the noise
`
`estimate Nk when the minimum estimate Mk is less than Nk as specified by claim 8.
`
`Resp., 41-45. In response, Andrea asserts that Apple double counts Mk to satisfy
`
`both the “future minimum” value and the “current magnitude value” of claim 8.
`
`
`
`17
`
`

`

`IPR2017-00627
`
`
`
`Petitioner’s Reply
`
`Andrea is incorrec