`
`Exhibit G
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 2 of 12
`
`IN THE UNITED STATES DISTRICT COURT
`FOR THE WESTERN DISTRICT OF TEXAS
`WACO DIVISION
`
`JAWBONE INNOVATIONS, LLC,
`
`Plaintiff,
`
`V.
`
`META PLATFORMS, INC.,
`D/B/AMETA,
`
`Defendant.
`
`Case No. 6:23-cv-00158-ADA-DTG
`
`JURY TRIAL DEMANDED
`
`SUPPLEMENTAL DECLARATION OF CLIFF READER, PH.D.
`
`Dated: May 8, 2024
`
`I} (a I
`-u·fl~LJDL,
`
`CliffR.eader, Ph.D.
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 3 of 12
`
`
`
`I.
`
`INTRODUCTION
`
`1.
`
`2.
`
`My name is Cliff Reader, Ph.D. I am over 18 years of age and, if I am called upon to do so, I
`
`would be competent to testify as to the matters discussed below.
`
`I have prepared this declaration at the request of Defendant Meta Platforms, Inc. (“Meta”). This
`
`declaration is a supplement to a declaration dated April 2, 2024 that I previously submitted in
`
`this case, as well as an original declaration dated May 25, 2022 in another case concerning the
`
`same patents. Both declarations, including their exhibits and appendices, are incorporated by
`
`reference herein. I have not been asked by Meta to provide any additional opinions regarding
`
`any terms other than the terms that I discuss below.
`
`3.
`
`In this declaration, I give my opinions regarding the view of a person of ordinary skill in the art
`
`of certain terms in the claims of the Asserted Patents. This declaration is based on information
`
`currently available to me, and I am willing to testify on the topics addressed below. This case is
`
`ongoing, and I may supplement or amend these opinions based on the results of further analysis
`
`and in rebuttal to positions taken by the Plaintiff. Because this declaration is based on
`
`information currently available to me, I reserve the right to continue my investigation, to
`
`review documents and information that may be produced, and to consider declarations,
`
`briefing, and deposition testimony from future depositions in this case. Therefore, I reserve the
`
`right to supplement, expand, and/or modify my opinions as my investigation continues and in
`
`response to any additional information that comes to my attention, including matters raised by
`
`the Plaintiff and other opinions provided by the Plaintiff’s expert(s).
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 4 of 12
`
`II. MATERIALS AND OTHER INFORMATION CONSIDERED
`
`4.
`
`I have reviewed each of the Asserted Patents along with their file histories, the parties’
`
`exchange of terms and proposed constructions, Jawbone’s Responsive Claim Construction
`
`Brief and Dr. Sayeed’s accompanying declaration, both dated April 24, 2024, and the evidence
`
`identified therein. I have also considered the sources cited below as well as my over 45 years of
`
`educational and professional experience in research and development in the areas of
`
`engineering, video/audio compression, audio/video transmission, signal and audio processing,
`
`real-time processing and display, system and algorithm design, and communications. In
`
`addition to the documents cited within this declaration, materials I considered are listed in
`
`Exhibit C of my April 2 2024 declaration.
`
`III. DISPUTED TERMS
`
`5.
`
`6.
`
`In this section, I supplement my opinions in my April 2, 2024 declaration regarding the view of
`
`a POSITA of certain terms in the asserted claims of the Asserted Patents.
`
`I have reviewed the ’213, ’611, ’080, ’691, and ’357 patents and their corresponding file
`
`histories and have confirmed that for each patent, the terms “approximately similar” and
`
`“approximately dissimilar,” or “substantially similar” and “substantially dissimilar” were
`
`recited in the claims as originally filed, and were not introduced through amendment.
`
`Moreover, as I explained in my declaration dated April 2, 2024 submitted in this case, none of
`
`the patents (including the claims and specification) provides any objective boundaries that are
`
`applicable to the terms “approximately similar” noise responses and “approximately dissimilar”
`
`speech responses, or “substantially similar” noise responses and “substantially dissimilar”
`
`speech responses. Thus, it is my opinion that the POSITA reading the patents and prosecution
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 5 of 12
`
`histories would not have been able to determine any applicable objective boundaries to these
`
`terms.
`
`7.
`
`I understand that Jawbone and Dr. Sayeed contend that the POSITA would be able to apply
`
`these terms without construction “because the POSITA would know that the noise responses
`
`must be similar enough, and the speech responses must be dissimilar enough, that the noise
`
`signal can be accurately compared to the main signal to detect speech, even in noisy
`
`environments” and “that the noise signal can be subtracted from the main signal to result in
`
`cleaned and denoised speech in later processing.” Sayeed Decl. ¶ 60. I disagree.
`
`8.
`
`As an initial matter, the patents and claims make clear that the terms “approximately similar,”
`
`“approximate dissimilar,” “substantially similar,” and “substantially dissimilar” relate to how
`
`dissimilar speech responses are or how similar noise responses are. As I explain below, the
`
`patents do not tie these terms to speech detection or denoising, which are actually tied to other
`
`terms recited in the claims.
`
`9.
`
`Even if the POSITA would have known to that these terms were tied to speech detection or
`
`denoised speech, there are no objective boundaries for these terms in the patents or prosecution
`
`histories. Thus, it is my opinion that the POSITA would not have been able to apply these
`
`terms without construction.
`
`A.
`
`The ’213 and ’611 patents: The terms “approximately similar” and
`“approximately dissimilar” are indefinite.
`
`10.
`
`First, the POSITA would have understood that “approximately,” “similar,” and “dissimilar”
`
`are terms of degree, and are subjective words with imprecise and inexact boundaries. Because
`
`the word “approximately” means “roughly or in the region of” (see Collins English Dictionary
`
`Excerpt at 4), its boundary is inexact. The word “similar” means “showing resemblance in
`
`qualities, characteristics, or appearance; alike but not identical” (see id. at 6). For responses to
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 6 of 12
`
`speech and responses to noise, there are various qualities, characteristics, and features of
`
`appearance that describe such responses. These may include, for example, the response
`
`amplitudes, slopes, and/or shapes. The responses can also vary across different parameters,
`
`such as frequency, direction, distance, or position of a sound source. Thus, when describing
`
`the overall shape or slope of a response, that could be defined across any such parameters. In
`
`my opinion, the POSITA would not understand which qualities, characteristics, or features of
`
`appearance to assess, nor the degree to which they must be similar or dissimilar, in order to
`
`meet the terms “approximately dissimilar” or “approximately similar.”
`
`11.
`
`Second, Jawbone and Dr. Sayeed rely heavily on a 10 dB teaching in the patents. The relevant
`
`sentence in the patents is: “When 0.8<B<1.1, the V1/V2 ratio is above approximately 10 dB—
`
`enough for good performance.” But this portion of the patents (and the surrounding portions)
`
`does not describe 10 dB as a standard for measuring similarity between two virtual
`
`microphones, let alone “approximately” so. Fundamentally, there is nothing in the patents or
`
`prosecution histories that say how similar should the noise responses be so that they are
`
`“approximately similar.” Nor do they say how dissimilar should the speech responses be so
`
`that they are “approximately dissimilar.” In fact, the patent drafter appears to have selected the
`
`10 dB threshold arbitrarily, and there is no explanation in the patent for why 10 dB was
`
`selected, as opposed to another threshold such as 5 dB or 15 dB.
`
`12.
`
`As the relevant sentence quoted above makes clear, the patents’ 10 dB teaching is “enough for
`
`good performance.” However, in my opinion, that “good performance” is unrelated to
`
`determining similarity, and there is no reason for the POSITA to insert “good performance” as
`
`a requirement in the claim. This is especially true for these patents, which at various points
`
`discuss different levels of performance, including “adequate performance” using 3 dB. See,
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 7 of 12
`
`e.g., ’213 patent at 20:23-26. The POSITA would have no reason to reject the patents’ 3 dB
`
`teaching in favor of the 10 dB teaching. The patents also describe additional levels of
`
`performance, including “nominal performance,” “ultimate performance,” “well-performing
`
`system,” and systems that “perform well,” for “excellent noise suppression performance,” and
`
`“optimal performance.” See id. at 23:52-53, 28:18-20, 20:8-37, 10:1-2, 27:67-28:1, 28:15-17,
`
`29:65-30:3. Thus, even assuming that some level of performance should be required by the
`
`claims, the patents describe various levels of performance, and the POSITA would be uncertain
`
`as to whether the claims require performance that is “good” versus one or more of the other
`
`levels of performance articulated in the patents.
`
`13.
`
`In fact, nowhere do the patents describe “approximately similar” noise responses and
`
`“approximately dissimilar” speech responses as being associated with any specific level of
`
`performance. Instead, where the patents do discuss performance with responses together, they
`
`discuss a system “performs well,” the responses are “very different” or “much less similar.”
`
`See, e.g., ’213 patent at 29:59-30:3 (stating “the system still performs well” when “very
`
`different speech responses remain” and “the noise responses are much less similar, so
`
`denoising will not be as effective.”); 29:40-43 (stating that the “formulations for V1 and V2
`
`can be varied and still result in good performance of the system as a whole.”). A POSITA
`
`would not consider “approximately dissimilar” speech responses to be the same as “very
`
`different” speech responses. Nor would a POSITA consider “approximately similar” noise
`
`responses to be the same as “much less similar” noise responses.
`
`14.
`
`Third, at most, the 10 dB teaching is for speech responses, not noise responses, so it cannot
`
`apply to at least the “approximately similar” noise responses limitation. This is made clear in
`
`the text, which refers the ratio V1/V2 as the “ratio of speech responses.” ’213 patent, 26:45-58.
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 8 of 12
`
`There is nothing in the patent that mentions a 10 dB threshold for determining whether two
`
`noise responses are “approximately similar” or not. Therefore, it is my opinion that even
`
`applying the 10 dB teaching to the claims, the 10 dB would only measure the difference
`
`between speech responses, and the POSITA would still be unable to differentiate between two
`
`noise responses that are “approximately similar” versus noise responses that are not
`
`“approximately similar.”
`
`15.
`
`In order to apply the 10 dB teaching to the noise responses, Jawbone and Dr. Sayeed first
`
`allege that any two responses that are not “approximately similar” must be “approximately
`
`dissimilar.” I disagree. The POSITA would not consider that two responses must be only one
`
`of “approximately similar” and “approximately dissimilar.” For example, there could be
`
`responses that are neither “approximately similar” and “approximately dissimilar,” or both,
`
`such as when the ranges for these terms overlap. This is especially true given the use of the
`
`word “approximately,” which is an elusive term of degree.
`
`16.
`
`Jawbone and Dr. Sayeed next provide a lengthy derivation of 9.54 dB as a threshold for
`
`defining whether two noise responses are “approximately similar.” However, to arrive at this
`
`value, Dr. Sayeed’s algebraic derivation is full of mathematical assumptions and logical leaps,
`
`none of which the POSITA would have understood to apply.
`
`17.
`
`As one example, the derivation is based on a combination of embodiments in the patent, first
`
`relying on an “optimal” system, and then injecting mathematical assumptions (including that β
`
`must equal 0.8 for noise, and that the B values must vary within a specific range of 0.8 to 1.1
`
`for noise) into a separate “non-optimal” system.
`
`18.
`
`As another example, Dr. Sayeed assumes that noise is only taken in the forward direction.
`
`However, I understand that Jawbone (and indeed Dr. Sayeed) has previously represented to the
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 9 of 12
`
`Patent Trial and Appeal Board (PTAB) that noise actually comes from all different directions,
`
`not just the forward direction. See Patent Owner Preliminary Responses and accompanying
`
`declarations of Dr. Sayeed in IPR2023-01130, IPR2024-00349, and IPR2024-00350. Noise
`
`comes from the direction where the noise source is positioned, which can be from any
`
`direction. Thus, the assumption that noise can only come from the forward direction is
`
`incorrect.
`
`19.
`
`In addition, Dr. Sayeed’s analysis for deriving 9.54 dB again requires there to be “good
`
`performance,” which the POSITA would not have known to apply for the reasons discussed
`
`above.
`
`20. Moreover, Dr. Sayeed’s analysis assumes that the values for B1 and B2 must be confined to
`
`and independently varied within a specific range of 0.8 to 1.1. This logic is circular, as the
`
`patents rely on the 10 dB threshold in order to identify the suitable range of values for B as 0.8
`
`to 1.1. Now, Dr. Sayeed relies on this range of B values of 0.8 to 1.1, in order to arrive at the
`
`same 10 dB value. In my opinion, there should be no dispute that the patents are completely
`
`silent as to any specific boundaries, let alone a 9.54 dB threshold, for “approximately similar”
`
`noise responses.
`
`21.
`
`Given Jawbone’s and Dr. Sayeed’s statements in the PTAB proceedings that the noise
`
`responses include different directions, this makes measuring the similarity between two noise
`
`responses even more uncertain. There is nothing in the patents (or Jawbone’s proposed
`
`constructions) that explain whether, in order to be “approximately similar,” a 10 dB threshold
`
`should apply to noise responses for the same direction or different directions, or in all
`
`directions. Jawbone and Dr. Sayeed do not address this point.
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 10 of 12
`
`22.
`
`Finally, the POSITA would not have understood that subtracting noise responses that are nearly
`
`10 dB apart results in denoising. Even assuming Jawbone and Dr. Sayeed are correct that
`
`denoising should occur without devoicing, a 10 dB threshold for “approximately similar” noise
`
`responses does not make sense. Jawbone’s and Dr. Sayeed’s position is that two virtual
`
`microphone speech responses that are 10 dB are approximately dissimilar, yet two virtual
`
`microphone noise responses that are nearly 10 dB (e.g., 9.99 dB) are approximately similar
`
`(because 9.99 has an absolute value of less than 10 dB and meets Jawbone’s and Dr. Sayeed’s
`
`construction). Somehow, Jawbone’s position is that the same threshold should apply to speech
`
`responses to be “approximately dissimilar” AND noise responses to be “approximately
`
`similar.” That is, according to Jawbone’s constructions, when the speech responses are
`
`subtracted (resulting in a difference of 10 dB), devoicing does not occur and speech is not
`
`cancelled out. But somehow, when the noise responses are subtracted (resulting in a difference
`
`of 9.99 dB), denoising does occur and noise is cancelled out. It would be unclear to a POSITA
`
`how noise responses that are nearly 10 dB apart would result in good noise cancellation,
`
`especially if responses that are 10 dB apart is supposed to avoid cancellation.
`
`23.
`
`Fourth, the claims of the ’213 and ’611 patents additionally recite detecting “acoustic voice
`
`activity,” and specifically require determining that acoustic voice activity is present “when an
`
`energy ratio of energies of the first virtual microphone and the second virtual microphone is
`
`greater than a threshold value.” ’213 patent at claims 1, 14, 42; see also ’611 patent at claims
`
`1, 29, 44 (reciting similar language). In my opinion, at most, the POSITA would interpret the
`
`patents’ 10 dB teaching (“V1/V2 ratio is above approximately 10 dB”) to be tied to the claimed
`
`“energy ratio” being “greater than a threshold value,” not the terms in dispute. It is my
`
`understanding that when a patent uses two different terms in the claims, there is a presumption
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 11 of 12
`
`that the two terms have different meanings. Thus, the terms in dispute must have a different
`
`meaning from “energy ratio” and “threshold value.” This means that the patents’ 10 dB
`
`teachings cannot also provide a measure for whether the noise responses are “approximately
`
`similar” or whether the speech responses are “approximately dissimilar.”
`
`B.
`
`The ‘080, ‘691, and ‘357 patents: The term “substantially similar” and
`“substantially dissimilar” are indefinite.
`
`24.
`
`I incorporate the paragraphs above for the ’213 and ’611 patents for the “substantially similar”
`
`and “substantially dissimilar” terms. The patents are similar and have the same specific
`
`relevant teachings that I rely on above for the “approximately similar” and “approximately
`
`dissimilar” terms. See, e.g., ’080 patent, 6:31-34, 6:47-62, 10:15-17, 14:28-50, 16:15-31.
`
`25.
`
`In my opinion, the 10 dB teaching cannot apply to the disputed terms in the ’080 or ’357
`
`patents. Even accepting an interpretation that the claims require “good performance” of
`
`“sufficient denoising and minimal devoicing,” that requirement would apply to other claim
`
`limitations, not the terms in dispute. Specifically, the ’080 patent claims separately recite an
`
`“adaptive noise removal application” that provides denoised output signals (see claims 1, 7,
`
`14), and the ’357 patent claims separately recite an output signal having noise content that is
`
`attenuated with respect to speech content (see claims 1, 15). If anything, the POSITA would
`
`have understood the patents’ 10 dB teaching to apply to these limitations, not “substantially
`
`dissimilar,” and certainly not “substantially similar” for the same reasons discussed above for
`
`“approximately similar.” Moreover, the 10 dB teaching does not apply to the ’691 patent,
`
`which merely claims microphone arrays. The claims do not require any noise removal or voice
`
`activity detection, nor do they require that the two virtual microphones be used to cancel noise
`
`or to avoid devoicing. The POSITA would not have understood these claims to require the
`
`
`
`
`
`
`Case 6:23-cv-00158-ADA Document 76-7 Filed 05/08/24 Page 12 of 12
`
`patent’s specific 10 dB teaching for “good performance” of “sufficient denoising and minimal
`
`devoicing.”
`
`26.
`
`Finally, Jawbone and Dr. Sayeed allege that for claims 7 and 14 of the ’080 patent,
`
`“substantially similar” means “within a range of 10 dB [across a plurality of frequencies for a
`
`speech source]”. Jawbone and Dr. Sayeed point to Fig. 13 of the ’080 patent, but the only
`
`teaching there is that the response “is flat for all frequencies within approximately +-30 degrees
`
`of the axis of the array.” ’080 Patent at 11:43-49. All this teaches the POSITA is that the
`
`responses for different frequencies are “flat.” As depicted in the figure, the “flat” responses
`
`means that the responses are all the same within the range of angles from -30 to +30.
`
`Responses that are all the same have a 0 dB difference, which is nowhere close to 10 dB. In
`
`short, there is nothing that ties 10 dB to the “flat” teaching in the ’080 patent, let alone the ’080
`
`patent claim requirement for the response to be “substantially similar” across frequencies. The
`
`POSITA would not understand Fig. 13 to support a threshold of 10 dB for “substantially
`
`similar” across frequencies.
`
`
`
`
`
`
`