`
`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
`____________________
`
`IPR2017-00626
`__________________________________________________________________
`
`Petitioner’s Opening Remand Brief
`
`
`
`IPR2017-00626
`
`
`
`
`
`Table of Contents
`
`I.
`II.
`
`Introduction .................................................................................................... 1
`Procedural Background ................................................................................ 1
`A.
`Summary of the Claims at Issue ........................................................ 1
`B.
`Scope of Issues to Be Decided ............................................................. 2
`III. Argument ........................................................................................................ 3
`A. Martin’s Noise Floor Tracking Algorithm Teaches the Elements
`of Claims 6-9 ........................................................................................ 3
`B. Martin Teaches the Future Minimum of Claim 4 ............................ 4
`C. Martin Teaches Setting a Current Minimum to a Future
`Minimum Value “Periodically,” as Required by Claim 6 ............... 8
`The Skilled Person Would Have Been Motivated to Set Hirsch’s
`Noise Threshold Using Martin’s Noise Tracking Algorithm .......... 9
`IV. Conclusion .................................................................................................... 10
`
`
`
`D.
`
`
`
`ii
`
`
`
`IPR2017-00626
`
`
`
`
`
`TABLE OF AUTHORITIES
`
`Cases
`KSR Intern. Co. v. Teleflex Inc.,
`127 S. Ct. 1727 (2007) ........................................................................................ 10
`
`Page(s)
`
`
`
`
`
`iii
`
`
`
`IPR2017-00626
`
`
`
`
`
`I.
`
`Introduction
`The sole question to be addressed in this remanded proceeding—whether
`
`claims 6 to 9 of the ’345 Patent would have been obvious over Hirsch (Ex. 1005)
`
`and Martin (Ex. 1006)—has been substantially narrowed by prior events in this
`
`case. First, Andrea did not appeal the Board’s finding that the apparatus described
`
`in Hirsch meets every requirement of claim 1 of the ’345 Patent. Second, Andrea
`
`did not appeal the Board’s determination that claim 25 was obvious in view of
`
`Hirsch and Martin, and now cannot dispute that a skilled person would have
`
`considered Hirsch and Martin together or that “techniques such as those shown in
`
`Hirsch and Martin are routinely combined.” Final Written Decision (FWD) at 16;
`
`see also 949 F.3d 697, 703 (Fed. Cir. 2020) (“Hirsch refers to Martin as a ‘known’
`
`approach ‘to avoid the problems of speech pause detection and to estimate the
`
`noise characteristics just from a past segment of noisy speech.”). The obviousness
`
`question thus distills down to a question of how Martin’s sound floor algorithm
`
`works, and whether when integrated into Hirsch, the resulting device will meet the
`
`requirements of claims 6 to 9. The answer, based on the arguments and evidence
`
`in this record, is yes. The Board should find claims 6 to 9 unpatentable.
`
`II.
`
`Procedural Background
`Summary of the Claims at Issue
`A.
`Claims 6-9 ultimately depend from claim 1, which defines “[a]n apparatus
`
`for canceling noise” by, inter alia, setting a threshold for each frequency bin using
`
`
`
`1
`
`
`
`IPR2017-00626
`
`
`
`
`
`a noise estimation process and detecting whether each bin’s magnitude is less than
`
`the corresponding threshold. Ex. 1001, 9:34-46. Claim 4 depends from claim 1
`
`and specifies “…set[ting] the threshold for each frequency bin in accordance with
`
`a current minimum value of the magnitude of the corresponding frequency bin;
`
`said current minimum value being derived in accordance with a future minimum
`
`value of the magnitude of the corresponding frequency bin.” Id., 9:54-60. Claim 5
`
`adds that the “future minimum value is determined as the minimum value of the
`
`magnitude of the corresponding frequency bin within a predetermined period of
`
`time.” Claim 6 depends from claim 5 and specifies that “said current minimum
`
`value is set to said future minimum value periodically.”1 Id., 9:65-67. Claims 7
`
`and 8 depend from claim 6, and address updating of the future minimum and
`
`current minimum values, respectively, while claim 9 depends from claim 5, and
`
`addresses updating of the future minimum value.
`
`Scope of Issues to Be Decided
`B.
`The Board’s remand order (Paper 28) specifies that briefing is limited to
`
`issues previously raised in the Response, Reply, or Observations on Cross. In
`
`those papers, Andrea addressed Apple’s patentability challenges to claims 6-9 by
`
`contending that Hirsch and Martin did not teach an apparatus as defined in claims
`
`
`
`1 “Periodically” means a fixed interval of time. See 949 F.3d at 709.
`
`
`
`2
`
`
`
`IPR2017-00626
`
`
`
`
`
`4 and 6.2 As Andrea did not present separate arguments regarding claims 5 and 7-
`
`9, it should not be permitted to do so now. See Reply, 3 n.2.
`
`Andrea also did not appeal several of the Board’s findings in its Final
`
`Written Decision which are now finally decided and cannot be reconsidered.
`
`Those include the Board’s finding that a skilled person would have considered
`
`Hirsch and Martin together. See FWD at 16 (citing Apple’s explanations that
`
`“‘because Hirsch not only cites to Martin, but identifies relevant benefits it
`
`provides,’ and ‘[a] skilled person would also have recognized that techniques such
`
`as those shown in both Hirsch and Martin are routinely combined.’”). Andrea thus
`
`cannot now dispute that a skilled person would have considered the combined
`
`teachings of Hirsch and Martin and that such teachings were routinely combined.
`
`III. Argument
`A. Martin’s Noise Floor Tracking Algorithm Satisfies Claims 6-9
`Andrea has maintained in these proceedings that Martin does not describe or
`
`suggest (i) the “future minimum” element of claim 4, from which claims 6-9
`
`depend or (ii) the future minimum setting requirement of claim 6. The full record
`
`of this proceeding refutes both of Andrea’s contentions.
`
`
`
`2 Andrea did not appeal the Board’s finding in IPR2017-00627 that claims 4-5 and
`
`10-11 are unpatentable.
`
`
`
`3
`
`
`
`IPR2017-00626
`
`
`
`
`
`1. Martin Teaches the Future Minimum of Claim 4
`As Apple explained in the original briefing, Martin teaches a noise
`
`estimation algorithm that determines the noise floor of an audio signal by tracking
`
`its minimum value within successive windows of time. In Martin’s technique,
`
`each window has a defined duration (e.g., of 0.625 seconds) and is comprised of L
`
`digital samples. Pet., 33, 40; Ex. 1003, ¶¶135-39; Ex. 1006 at 1093. Martin
`
`explains that each window can be divided into an arbitrary number of sub-windows
`
`W, each of the same length M (i.e., L = M x W), and illustrates its algorithm with an
`
`example using 4 sub-windows (W=4). Ex. 1003, ¶¶135, 139; Ex. 1023, ¶20.
`
`Martin uses the variable PMmin to track the signal’s minimum value within a
`
`sub-window. This variable is the “future minimum” of the claims. Pet., 38-39, 42.
`
`At the end of a sub-window, PMmin is
`
`used to update the noise floor Pn(i). The
`
`variable Pn(i) is the “current minimum.”
`
`Martin describes two ways of
`
`updating the noise floor based on the
`
`nature of the signal in the window of
`
`samples L. If the signal is monotonically
`
`increasing (i.e., its minimum value
`
`increases across successive sub-
`
`
`
`4
`
`
`
`IPR2017-00626
`
`
`
`
`
`windows), then Martin’s algorithm sets Pn(i) based on the value of PMmin observed
`
`in each sub-window. That has the effect of setting the noise floor Pn(i) (“current
`
`minimum”) based on the value of PMmin (“future minimium”) in the last sub-
`
`window of the window (e.g., W=4). If the signal is not monotonically increasing,
`
`Martin’s algorithm sets Pn(i) (“current minimum”) to the smallest PMmin (“future
`
`minimum”) value observed in the entire window of L samples (e.g., across the 4
`
`sub-windows in Martin’s illustration). See Ex. 1006, 1094; Ex. 1003, ¶¶135-36;
`
`Ex. 1023, ¶¶16-20. Martin also makes clear that signal characteristics dictate
`
`which option is used—if the PMmin value over the past W sub-windows is
`
`monotonically increasing, option (1) is used, otherwise option (2) is used. Ex.
`
`1006, 1094; Fig. 2. Martin’s algorithm thus tracks a current minimum (i.e., Pn(i))
`
`and derives it in accordance with a future minimum (PMmin) as claim 4 specifies.
`
`Andrea previously contended that Martin does not teach claim 4’s “future
`
`minimum value” because PMmin is the minimum of a sub-window and not of an
`
`entire window. Resp., 25. This, according to Andrea, means that Martin’s
`
`algorithm supposedly does not set the current minimum to the lowest observed
`
`value within a window of L samples when the sound is monotonically increasing.
`
`Id. Andrea also contended that when the sound is not monotonically increasing,
`
`Martin’s algorithm sets the future minimum value in a different way than its
`
`putative invention. Resp., 26-27.
`
`
`
`5
`
`
`
`IPR2017-00626
`
`
`
`
`
`Initially, neither of Andrea’s supposed distinctions are based on the claims.
`
`Nothing in claims 6 to 9 requires the sample being processed to have any minimum
`
`duration, and nothing in the claims imposes any requirements on how the future
`
`minimum value is derived from a current minimum.
`
`Andrea also ignores that Martin explicitly teaches what Andrea says it does
`
`not—that if the signal is not monotonically increasing, then Martin’s algorithm
`
`will set the Pn(i) variable (the “current minimum”) to the minimum value in the
`
`entire window of samples L. Ex. 1023, ¶¶19-20. As Martin states: “[i]n case of
`
`non monotonic power the noise power estimate is set to the minimum of the
`
`length L window...” Ex. 1006, 1094. Andrea’s only response here is its assertion
`
`that Martin uses a variable, “min_vec”, to temporarily store the PMmin values of
`
`each subwindow. See Resp., 27 (arguing that the noise floor Pn(i) is never set
`
`equal to PMmin because it is set to min_vec instead). But Andrea ignores the role of
`
`min_vec in Martin’s algorithm—that it is used temporarily to store the PMmin values
`
`of each sub-window, which enables the Martin algorithm to determine what the
`
`lowest value is within the entire window L of samples, and to thereby set the noise
`
`floor Pn(i) to the smallest PMmin value for that entire window of samples L.3 Ex.
`
`1006, 1094 (showing “min_vec(r*M)=PMmin”); Ex. 1023, ¶19.
`
`
`
`3 If W=4, Martin will set Pn(i) to the smallest of the past 4 PMmin values in min_vec.
`
`
`
`6
`
`
`
`IPR2017-00626
`
`
`
`
`
`Andrea also cannot dispute this is what Martin’s algorithm does. Andrea’s
`
`expert, Dr. Douglas confirmed at his deposition that the min_vec variable stores the
`
`PMmin values, and that Martin sets the noise floor Pn(i) to be equal to the smallest of
`
`those PMmin values:
`
`Q And the min vec values are the PMmin values, correct?
`A They are the PMmin values.
`Q So this assignment in Figure 2 of Martin assigns Pn(i) to the
`smallest PMmin in the data window, correct?
`A Yes.
`
`Ex. 1030, 177:5-16; see also Ex. 1026, 87:15-20, 103:7-104:12.
`
`Andrea also argued that, in the scenario where a signal is monotonically
`
`increasing, Martin only describes setting the noise floor Pn(i) to be the PMmin value
`
`of the most recent sub-window, which will not be the minimum of the entire
`
`window L, and therefore is not a “future minimum.” Resp., 25-26. But nothing in
`
`the claims requires use of the lowest value in the entire window L to derive the
`
`“future minimum”—claim 4 imposes no minimum duration of the sample, nor does
`
`it restrict how the determination is to be made. See Ex. 1023, ¶18. The only
`
`reference to time in the claims is found in claim 5, which depends from claim 4,
`
`and specifies simply that the “future minimum” is determined “over a
`
`predetermined period of time.” Because the sub-windows in Martin’s algorithm
`
`have a fixed interval of time, they satisfy that requirement. Ex. 1003, ¶¶138, 141.
`
`
`
`7
`
`
`
`IPR2017-00626
`
`
`
`
`
`Importantly, Andrea also has admitted that Martin’s variable PMmin tracks the
`
`minimum value during each sub-window, which is a predetermined period of M
`
`samples. Resp., 24 (“In the monotonically increasing case, PMmin will always
`
`represent the most recent subwindow minimum…”). That admission, coupled with
`
`the absence of any claim requirements on duration of the signal being processed, is
`
`fatal to Andrea’s argument. Nothing in claim 4 (or any other claim) requires use of
`
`a predetermined period of a particular length (e.g., L) or prohibits tracking the
`
`minimum of a sub-window. Thus, in the scenario where a signal is monotonically
`
`increasing, Martin also satisfies the claims. Ex. 1023, ¶18.
`
`B. Martin Teaches Setting a Current Minimum to a Future
`Minimum Value “Periodically,” as Required by Claim 6
`With respect to claim 6, Andrea contended that Martin does not teach
`
`updating the current minimum “periodically,” incorrectly asserting that Martin sets
`
`Pn(i) (“current minimum”) equal to PMmin (“future minimum”) only “sporadically.”
`
`Resp., 28. According to Andrea, Martin sets Pn(i) to PMmin only in the
`
`monotonically increasing scenario, and thus, because the signal sometimes
`
`monotonically increases and sometimes does not, this assignment occurs randomly
`
`not periodically. Id. Andrea’s argument, however, rests on its flawed position
`
`that, in the non-monotonically increasing case, Martin’s use of min_vec to
`
`determine the smallest PMmin values in a data window means that the noise floor
`
`Pn(i) is not set equal to a PMmin value. As explained above, Andrea’s distinction
`
`
`
`8
`
`
`
`IPR2017-00626
`
`
`
`
`
`between min_vec and PMmin is illusory, as Dr. Douglas admitted. Ex. 1030, 177:5-
`
`16 (“Q So this assignment in Figure 2 of Martin assigns Pn(i) to the smallest
`
`PMmin in the data window, correct? A Yes.”); Ex. 1026, 87:15-20, 103:7-104:12.
`
`What Martin shows instead is that it updates the value Pn(i) at a fixed
`
`interval—after every M samples, as Dr. Douglas admitted. Ex. 1026, 77:16-80:19.
`
`Thus, whether the signal is monotonically increasing or not, Pn(i) is set to PMmin at
`
`the end of every sub-window of M samples, as shown in the blue box of annotated
`
`Figure 2 of Martin (reproduced above). Ex. 1023, ¶20; Ex. 1003, ¶¶138, 141.
`
`Because Martin sets Pn(i) equal to PMmin every M samples, it does so “at regular
`
`intervals of time” consistent with the Federal Circuit’s construction of
`
`“periodically” in claim 6.
`
`C. The Skilled Person Would Have Been Motivated to Set Hirsch’s
`Noise Threshold Using Martin’s Noise Tracking Algorithm
`Given the Board’s findings supporting its determination that claim 25 was
`
`obvious and Andrea’s failure to appeal that decision, Andrea cannot now dispute
`
`that a skilled person would have considered Hirsch and Martin together, or that
`
`“techniques such as those shown in Hirsch and Martin are routinely combined.”
`
`FWD at 16. Those findings, moreover, rest on substantial evidence and Apple’s
`
`prior arguments. See, e.g., Pet., 34-38 and Ex. 1003, ¶¶125-32 (explaining the
`
`
`
`9
`
`
`
`IPR2017-00626
`
`
`
`
`
`skilled person would have found it obvious to modify Hirsch’s adaptive threshold4
`
`calculation “to use Martin’s noise floor algorithm instead of Hirsch’s running
`
`average algorithm” as it would improve operation of Hirsch in non-stationary
`
`environments and because Hirsch cites Martin as a known technique); Ex. 1003,
`
`¶¶131-32 (in audio signal processing, “[i]t was standard to attempt to optimize
`
`system performance by swapping algorithms or tuning parameters”); Ex. 2005,
`
`129:1-23 (a skilled artisan would have been motivated to try adding Martin’s
`
`algorithm to Hirsch to determine whether it performed better); see also, e.g., KSR
`
`Intern. Co. v. Teleflex Inc., 127 S. Ct. 1727, 1740 (2007) (“…if a technique has
`
`been used to improve one device, and a [POSA] would recognize that it would
`
`improve similar devices in the same way, using the technique is obvious unless its
`
`actual application is beyond his or her skill.”). As explained above, modifying
`
`Hirsch’s adaptive threshold to use Martin’s noise floor algorithm would yield a
`
`system that meets claim 4’s requirement of “set[ting] the threshold… in
`
`accordance with a current minimum”). Ex. 1003, ¶127; Pet., 36-37, 47.
`
`IV. Conclusion
`Apple respectfully submits that the Board should cancel claims 6-9.
`
`
`
`4 Hirsch uses the adaptive threshold to distinguish between speech and noise;
`
`signals above the threshold are speech, below are noise.
`
`
`
`10
`
`
`
`IPR2017-00626
`
`
`
`
`
`
`
`Dated: April 22, 2020
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Respectfully Submitted,
`
`/Jeffrey P. Kushan/
`Jeffrey P. Kushan
`Registration No. 43,401
`Sidley Austin LLP
`1501 K Street NW
`Washington, DC 20005
`jkushan@sidley.com
`(202) 736-8914
`
`Lead Counsel for Petitioner
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`11
`
`
`
`IPR2017-00626
`
`
`
`
`
`CERTIFICATE OF SERVICE
`
`I hereby certify that on this 22nd day of April, 2020, copies of this
`
`Petitioner’s Opening Remand Brief, and Exhibits have been served in its entirety
`
`by email on the following counsel of record for Patent Owner:
`
`
`William D. Belanger, belangerw@pepperlaw.com
`Frank D. Liu, liuf@pepperlaw.com
`Andrew P. Zappia, zappiaa@pepperlaw.com
`BN_IPR-Andrea@pepperlaw.com
`
`Respectfully submitted,
`
`/Jeffrey P. Kushan/
`Jeffrey P. Kushan
`Reg. No. 43,401
`Attorney for Petitioner
`
`
`
`
`Dated:
`
`April 22, 2020
`
`
`
`
`
`
`
`