`_________________
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`_________________
`MICROSOFT CORPORATION, MICROSOFT MOBILE INC., SAMSUNG
`ELECTRONICS AMERICA, INC. AND SAMSUNG ELECTRONICS CO. LTD.
`Petitioners
`
`v.
`
`FASTVDO LLC
`Patent Owner
`_________________
`U.S. Patent No. 5,850,482
`Issued: December 15, 1998
`Application No.: 08/633,896
`Filed: April 17, 1996
`Title: Error Resilient Method And Apparatus For Entropy Coding
`_________________
`DECLARATION OF DR. ROBERT L. STEVENSON IN SUPPORT
`OF PETITIONERS’ PETITION FOR INTER PARTES REVIEW OF
`CLAIMS 1-3, 5-6, 12-14, 16-17, AND 28 OF U.S. PATENT NO. 5,850,482
`
`Page 1 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`1.
`
`I, Dr. Robert L. Stevenson, declare that all statements made herein of
`
`my own knowledge are true and all statements made on information and belief are
`
`believed to be true; and further that these statements were made with the
`
`knowledge that willful false statements and the like so made are punishable by fine
`
`or imprisonment, or both, under Section 1001 of Title 18 of the United States
`
`Code.
`
`2.
`
`I have been hired by Klarquist Sparkman, LLP, counsel for Microsoft
`
`Mobile Inc. and Microsoft Corporation (“Microsoft”), and Samsung Electronics
`
`America, Inc. and Samsung Electronics Co. Ltd. (“Samsung,” collectively
`
`“Petitioners”) as an expert witness in the above-captioned proceeding (the “IPR”).
`
`I have been asked to provide my opinion regarding U.S. Patent No. 5,850,482 (“the
`
`482 patent”).
`
`I.
`
`BACKGROUND AND QUALIFICATIONS
`3. My Curriculum Vitae is attached to this Declaration as Exhibit A.
`
`A. Educational Background
`4.
`I have earned a Bachelor's degree in Electrical Engineering from the
`
`University of Delaware and a Ph.D. degree in Electrical Engineering from Purdue
`
`University. My Ph.D. research was on communications and signal processing.
`
`B.
`5.
`
`Professional History
`I am presently a Professor in the Department of Electrical Engineering
`
`and in the Department of Computer Science and Engineering at the University of
`
`1
`
`Page 2 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`Notre Dame. I first joined the faculty at the University of Notre Dame as an
`
`Assistant Professor in the Department of Electrical Engineering in 1990. I was
`
`granted tenure and promoted to the rank of Associate Professor in August 1996. I
`
`attained the rank of Professor in the Department of Electrical Engineering in
`
`August 2002, and I continue to serve in that capacity. I have served concurrently as
`
`a Professor in the Department of Computer Science and Engineering at the
`
`University of Notre Dame since January 2003.
`
`6.
`
`I spent the summer of 1992 at the Air Force Research Lab in Rome,
`
`New York and I spent the summer of 1993 at the Intel® Corporation in Hillsboro,
`
`Oregon. Several leading computing companies, including Intel®, Sun
`
`Microsystems®, and Apple® Computer have supported my research at Notre
`
`Dame. During the past 20 years, I have published over 150 technical papers
`
`related to the field of image processing and digital systems.
`
`7.
`
`I am a member of the Institute of Electronics and Electrical Engineers,
`
`The International Society for Optical Engineering, and the Society for Imaging
`
`Science and Technology. I am a member of the academic honor societies Eta
`
`Kappa Nu, Tau Beta Pi, and Phi Kappa Phi.
`
`8.
`
`For the past 20 years my work has focused on the design of
`
`techniques, hardware, and software for the processing of digital signals using
`
`digital computing devices. As an academic researcher I attempt to develop novel
`
`2
`
`Page 3 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`ideas for systems, then publish and present those ideas to the technical community.
`
`My success as an academic is directly related to the insights and techniques which
`
`provide the basis for new generations of products. My early work on digital
`
`techniques for printing and image capture devices led to significant interaction
`
`with companies developing desktop computers products in the early 1990's as they
`
`tried to incorporate those ideas into their products.
`
`9. My interaction with Apple's Imaging Group focused on various
`
`imaging devices such as digital cameras, scanners, and printers and how to best
`
`support those devices on desktop computers. At Intel, I worked in Intel's
`
`Architecture Lab at the time the MMX multimedia instructions were being
`
`incorporated into the Pentium processor. My work there dealt with developing
`
`video compression techniques for CD-ROM's and network communications that
`
`were well matched to the Pentium architecture. I also gave a series of talks on how
`
`advanced communication and video processing techniques could be better
`
`supported on the Pentium platform. Similarly, my interaction with Sun
`
`Microsystem's group examined how advanced signal processing techniques could
`
`be best implemented using Sun's new Visual Instruction Set on the Sparc
`
`architecture.
`
`10.
`
`I have also received significant support for my research from several
`
`U.S. Department of Defense Agencies. The Air Force Research Laboratory has
`
`3
`
`Page 4 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`funded my work to develop advanced parallel processing algorithms which
`
`exploited an ad-hoc network of mixed computers to achieve signification
`
`computational advantages over their previously implemented techniques. Other
`
`Department of Defense agencies have supported my work in image and video
`
`enhancement.
`
`11.
`
`I have published 33 journal articles, written 9 book chapters, edited
`
`the proceedings of 15 conferences, and presented 109 papers at professional
`
`conferences.
`
`12.
`
`I am an inventor of U.S. Patent No. 6,081,552, “Video Coding Using
`
`a Maximum A Posteriori Loop Filter,” June 27, 2000.
`
`13.
`
`If asked, I will testify regarding my qualifications, background and
`
`experience in the field of data compression, encoding and decoding.
`
`14.
`
`I am being compensated at a rate of $600 per hour for my study and
`
`testimony in this reexamination. I am also being reimbursed for reasonable and
`
`customary expenses associated with my work and testimony. My compensation is
`
`not contingent on the outcome of this Petition, the related litigation or the specifics
`
`of my testimony.
`
`II.
`
`STANDARDS
`15. As part of my work in connection with this matter, I have studied the
`
`482 patent, including the written description, figures, and claims. I have also
`
`4
`
`Page 5 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`reviewed the U.S. Patent and Trademark Office ("USPTO") file history of the 482
`
`patent, and certain of the parties’ infringement and invalidity contentions in the
`
`related litigation. I have also reviewed the prior art references cited in the Petition,
`
`as well as additional background references.
`
`16. My opinions are based on my years of education, research and
`
`experience, as well as my investigation and study of relevant materials. In forming
`
`my opinions, I have considered the materials referred to herein.
`
`17. Prior art is generally the state of technology in the relevant field at the
`
`time of the invention, including both systems described in publications, such as
`
`conference papers, and systems actually in use at some time prior to the patent
`
`filings. I understand that reexamination prior art is limited to “patents and printed
`
`publications.” I understand that claims in a reexamination are given their broadest
`
`reasonable interpretation in light of the specification.
`
`18.
`
`I understand that there are two ways in which a prior art patent or
`
`printed publication can be used to invalidate a patent. First, the prior art can be
`
`shown to “anticipate” the claim. Second, the prior art can be shown to “render
`
`obvious” the claim. My understanding of the obviousness legal standards is set
`
`forth below.
`
`19.
`
`I understand that an inventor is not entitled to a patent if his or her
`
`invention would have been obvious to a person of ordinary skill in the field of the
`
`5
`
`Page 6 of 49
`
`
`
`invention at the time the invention was made. I understand that 35 U.S.C. § 103(a)
`
`U.S. Patent No. 5,850,482
`
`states:
`
`A patent may not be obtained though the invention is not
`identically disclosed or described as set forth in section
`102 of this title, if the differences between the subject
`matter sought to be patented and the prior art are such
`that the subject matter as a whole would have been
`obvious at the time the invention was made to a person
`having ordinary skill in the art to which said subject
`matter pertains. Patentability shall not be negatived by
`the manner in which the invention was made.
`
`20. The following standards govern the determination of whether a claim
`
`in a patent is obvious. I have applied these standards in my evaluation of whether
`
`the claims in the 482 patent are obvious.
`
`21. A claim in a patent is obvious when the differences between the
`
`subject matter sought to be patented and the prior art are such that the subject
`
`matter as a whole would have been obvious at the time the invention was made to a
`
`person having ordinary skill in the art to which the subject matter pertains.
`
`22.
`
`I understand that obviousness may be shown by considering more
`
`than one item of prior art. I also understand that the relevant inquiry into
`
`obviousness requires consideration of four factors (although not necessarily in the
`
`following order):
`
` The scope and content of the prior art;
`
` The differences between the prior art and the claims at issue;
`
`6
`
`Page 7 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
` The level of ordinary skill in the pertinent art; and
`
` Whatever objective factors indicating obviousness or non-obviousness
`
`may be present in any particular case.
`
`23.
`
`In addition, I understand that the obviousness inquiry should be done
`
`through the eyes of a person of ordinary skill in the relevant art at the time the 482
`
`patent application was filed (in this case, April 17, 1996).
`
`OPINIONS REGARDING THE
`SUBJECT MATTER OF THE 482 PATENT
`III. BACKGROUND AND STATE OF THE ART
`24. The 482 patent (Ex. 1001)1 relates to methods and apparatus for
`
`compression using entropy encoding on quantized data. In particular the 482
`
`patent uses unequal levels of error coding for different subsets of a larger quantized
`
`data set. The background of the 482 patent admits that compression techniques
`
`such as transform-based compression were known to those skilled in the art, as
`
`were quantization techniques, entropy coding and decoding. The background also
`
`admits that using “Unequal Error Protection (UEP)” was also known to those of
`
`skill in the art. Nonetheless, the 482 patent claims the combination of these
`
`
`1 All Exhibit numbers in this Declaration refer to the Exhibits attached to the
`
`accompanying Petition for Inter Partes Review (the “Petition”).
`
`7
`
`Page 8 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`techniques into its claimed methods and its corresponding computer readable
`
`memory for performing those methods.
`
`25. There were a number of entropy coding techniques that were already
`
`well-known prior to the 482 patent application. As noted in the 482 patent, prior
`
`art entropy coding uses variable length coding to reduce the number of bits used to
`
`represent a data set. It accomplishes this by using variable length code words to
`
`represent symbols, with shorter code words for the most commonly occurring
`
`symbols. The most common entropy coding techniques are Huffman coding and
`
`arithmetic coding, with the main difference being that Huffman codes use an
`
`integer number of bits, while arithmetic codes can produce fractional numbers of
`
`bits.
`
`26. The patent acknowledges that all of this was already known to those
`
`of skill in the art. See Ex. 1001, 482 patent, 4:36-50.
`
`27. There were a number of known examples of transforming data before
`
`encoding prior to the 482 application, as the 482 patent states. Id., 2:11-25.
`
`28. Additionally, quantizing transformed data, including quantizing such
`
`data for subsequent entropy coding, was also well known, as the 482 patent also
`
`acknowledges. Id., 3:36-46.
`
`8
`
`Page 9 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`29. Also, the idea of applying unequal error protection (UEP) to subsets
`
`of data having unequal importance was also already well-known. See, e.g., id.,
`
`5:34-46.
`
`30. The 482 patent further acknowledges that splitting a code word into
`
`prefix and suffix fields was known in the art. Id., 16:28-32.
`
`31. The patent goes on to state that “the proposed codes have not
`
`previously been separated in order to provide error resiliency as provided by the
`
`method and apparatus of the present invention.” Ex. 1001, 482 patent, 16:33-35.
`
`As shown below in the discussion of the Kato prior art, this is simply not the case.
`
`IV. OVERVIEW OF THE 482 PATENT
`32.
`In the preferred embodiment, the 482 patent discusses employing
`
`known techniques for transforming and quantizing image data. This is done to
`
`provide fewer unique coefficients to represent the data before encoding, storing
`
`and/or transmitting the resulting quantized data using the techniques described
`
`below. See, generally, id., 9:30 – 13:3.
`
`33. A block diagram purporting to show an encoder implementing the
`
`claimed combination of an entropy encoder with unequal error protection is
`
`illustrated in Figure 2 of the 482 patent. Figure 2 includes steps for entropy
`
`encoding: a) “quantized coefficients using split field coding” (35) and b) “run
`
`length values” (37), and applying unequal error protection to the encoded data
`
`9
`
`Page 10 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`(38), wherein, e.g., “encoded run lengths” and “prefix fields” have “higher error
`
`protection” and “suffix fields” have “lower … or no error protection.”
`
`34. The 482 patent describes “split field coding” as the generation of a
`
`code word that includes a first portion (which the patent calls a “prefix field”) and
`
`a second portion (which the patent calls a “suffix field”). See id., 13:36-43.
`
`35. These prefix fields contain information “representative of a
`
`predetermined characteristic” of their corresponding suffix fields, including
`
`information representative of the length of the corresponding suffix field, e.g., the
`
`number of bits for the suffix field. See id., 7:18-25; see generally, id., Abstract;
`
`7:8-10, 15-17; 13:56-63; and 15:47-57.
`
`36. The 482 patent states that, as a result of this split field coding, even if
`
`the suffix fields are provided with a “lower level of error protection” (see, e.g., id.,
`
`7:30-40), the effect of any errors in a given suffix field will be minimized, and will
`
`not carry forward to other code words (assuming that the prefix field with which
`
`that suffix field is associated is decoded correctly). Id., 15:61-16:9.
`
`37. The 482 patent further describes an embodiment in which unequal
`
`error protection can be provided by storing the prefix fields using a higher level of
`
`error protection than is used (if any is used) for the suffix fields. Id., 482 patent,
`
`17:15-23.
`
`10
`
`Page 11 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`38. The 482 patent also describes an embodiment in which unequal error
`
`protection is provided by transmitting prefix fields over a first data link that is error
`
`protected, and by transmitting the suffix fields over a second data link that is not
`
`(or that is protected “to a lesser degree”). Id., 17:28-35.
`
`39. Thus, the 482 patent combines what it acknowledges are known
`
`techniques for entropy coding quantized image data (Ex. 1001, 482 patent, 4:36-
`
`39) with known “unequal error protection techniques” (id., 6:35-36) by using
`
`another known technique which the patent refers to as “split field coding” (id. at
`
`13:50). According to the patent, this “isolates the effects of a bit error to a single
`
`code word” (id., 6:38-39). However, this combination of admittedly prior art
`
`techniques was, itself, taught in the Kato reference.
`
`V. CLAIM CONSTRUCTION
`A. LEVEL OF ORDINARY SKILL IN THE ART
`40. Consistent with the background and standards set forth above, and in
`
`view of the prior art discussed herein, it is my belief that a person of ordinary skill
`
`in the relevant art of the 482 patent at the time the application that issued as the
`
`482 patent was filed (“POSA”) would have had an undergraduate degree in
`
`computer science, computer engineering, or electrical engineering, as well as two
`
`years’ experience or a graduate degree with focus in the area of data compression,
`
`encoding, and decoding. A POSA would have been familiar with and able to
`
`11
`
`Page 12 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`implement and recognize well-known data compression techniques. In particular,
`
`a POSA would have been familiar with techniques for both entropy coding and
`
`channel coding, including techniques for unequal error protection (UEP). A POSA
`
`would also have been able to recognize and implement variations of encoding
`
`techniques in the prior art. A POSA would have understood the different terms
`
`with which variations of these techniques had been explained. A POSA would
`
`also have been aware that various encoding techniques could be combined in
`
`parallel or in series and would have been able to adapt and combine well-known
`
`compression techniques.
`
`41.
`
`It is my understanding that the 482 patent expired on April 17, 2016,
`
`and therefore is not subject to amendment, and that based on that, for purposes of
`
`this Petition, the claims are construed pursuant to Phillips v. AWH Corp., 415 F.3d
`
`1303, 1312-13 (Fed. Cir. 2005) (words of a claim “are generally given their
`
`ordinary and customary meaning” as understood by a person of ordinary skill in
`
`the art in question at the time of the invention), whose qualifications I have
`
`discussed above.
`
`B.
`
`Terms That Appear In All Claims
` “code word”
`1.
`42. The term “code word” is present, either directly, or through
`
`dependence in all challenged claims. The 482 patent describes the process of
`
`12
`
`Page 13 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`assigning code words in entropy coding using a codebook: “The assignment of
`
`code words for entropy coding is typically governed by means of a codebook
`
`which must be known to both the encoder and decoder.” Ex. 1001, 482 patent,
`
`4:51-54 (emphasis added). The 482 patent further explains that according to the
`
`present invention, these code words represent “quantized image data.” Id., 13:36-
`
`41 (“According to the present invention, an entropy encoder 16 and, more
`
`preferably, code word generating means 26 generates a plurality of code words
`
`which are representative of the quantized significant coefficients. Accordingly, the
`
`plurality of code words effectively represent the quantized image data.”
`
`(Emphasis added.)) And, the patent describes entropy coding as using symbols to
`
`represent that information. Id., 14:27-31 (“As known to those skilled in the art,
`
`entropy coding achieves a reduction in the number of bits required to represent a
`
`data set by assigning shorter code words to symbols which occur frequently and
`
`longer code words to symbols which occur less frequently.” (Emphasis added.)).
`
`43. Accordingly, one of ordinary skill in the art would have understood
`
`“code word” in the 482 patent to mean: “code from a code book representing a
`
`symbol”.
`
`“first portion of each code word”
`2.
`44. The term “first portion of each code word” is present, either directly,
`
`or through dependence in all challenged claims. The patent expressly defines “first
`
`13
`
`Page 14 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`portion” as a “prefix field” (Ex. 1001, 482 patent, 13:41-43), and further explains
`
`that this first portion is “a first or prefix field which is susceptible to bit errors”
`
`(id., 6:52-55 (emphasis added)). As the patent further explains in the Summary of
`
`the Invention, “[T]he code words can be generated such that a bit error in the
`
`prefix field of a code word could result in a potential loss of code word
`
`synchronization.” Id., 6:56-59 (emphasis added).
`
`45. Accordingly, one of ordinary skill in the art would have understood
`
`“first portion of each code word” in the 482 patent to mean: “prefix field of a code
`
`word generated in a manner such that a bit error in the field could result in a
`
`potential loss of code word synchronization”.
`
` “[associated] second portion of each code word”
`3.
`46. The term “[associated] second portion of each code word” is present,
`
`either directly, or through dependence in all challenged claims. The patent
`
`expressly defines “[associated] second portion” as “an associated second or suffix
`
`field which is resilient to bit errors.” Ex. 1001, 482 patent, 6:52-56. The 482
`
`patent further explains what is meant by the phrase “resilient to bit errors.”
`
`Namely, the patent explains that
`
`In particular, the code words can be generated such that a bit
`error in the suffix field of a code word will not result in a loss
`of code word synchronization, but the resulting misdecoded
`value shall, instead, fall within a predetermined interval about
`the correct value.
`
`14
`
`Page 15 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`Id., 6:60-65 (emphasis added).
`
`47. Further, during prosecution, applicants explained that in the event of
`
`bit errors in the suffix field, “the code words are generated such that a bit error in
`
`the suffix field of a code word will not result in a loss of code word
`
`synchronization, but the resulting misdecoded value shall, instead, fall within a
`
`predetermined range about the correct value.” Ex. 1003, 482 file history, Jan. 23,
`
`1998 Amendment, p. 414 (emphasis added). In other words, one of ordinary skill
`
`in the art would understand that even in the presence of some decoding error in the
`
`second portion, the misdecoded value would be close to the actual value, with the
`
`value falling within some predetermined range of the actual value.
`
`48. Accordingly, one of ordinary skill in the art would have understood
`
`“[associated] second portion of each code word” in the 482 patent to mean:
`
`“[associated] suffix field of a code word generated in a manner such that a bit error
`
`in the field results in a miscoded value that falls in a predetermined range about the
`
`correct value.”
`
`VI. SUMMARY OF THE PRIOR ART TO THE 482 PATENT
`49. The attached Petition primarily relies on two pieces of prior art to
`
`challenge claims of the 482 patent, in addition to prior art and admissions showing
`
`the state of the art at the time of the 482 patent application’s filing.
`
`15
`
`Page 16 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`A. Kato Combines Entropy Coding, Split
`Field Coding, And Unequal Error Protection
`50. Kato states that its invention “relates to a method of efficient encoding
`
`which can reduce the total number of bits of recorded or transmitted data.” Ex.
`
`1002, Kato, 1:8-10. Kato describes that in “prior-art JPEG efficient encoding of
`
`the DCT (discrete cosine transform) type,” a frame represented by image data is
`
`divided into blocks of 8 x 8 pixels. This data is then subjected to a DCT transform,
`
`which generates a set of DCT coefficients, each of which includes one DC
`
`coefficient and 63 AC coefficients. Each of these DCT coefficients is quantized,
`
`and: “The resultant quantized AC coefficients are two-dimensionally encoded into
`
`a Huffman code. The resultant quantized DC coefficient is predictively encoded.”
`
`Id., 1:22-43.
`
`51.
`
`In predictive encoding, as Kato explains, based on the value for “the
`
`immediately-preceding input data,” an estimate is generated for a subsequent piece
`
`of data Di. Id., 1:55-57. This estimate or prediction Pi is generated based on
`
`previously encoded image data. When Pi is subtracted from the actual value for
`
`the input data Di, this generates what Kato refers to as an “estimation error” or
`
`“prediction error” Si, which is equal to the difference between the actual data Di
`
`and the prediction Pi. Id., 1:44-53. Because the difference values are usually
`
`smaller than the actual data values they can be encoded more compactly than the
`
`actual data values.
`
`16
`
`Page 17 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`52. Kato goes on to describe techniques for efficiently encoding this
`
`prediction error in an efficient manner. To do so, Kato teaches combining
`
`techniques for entropy coding and split field coding, and providing unequal error
`
`protection to the first and second portions into which given a code word is split.
`
`Kato Shows Combining Entropy Coding And Split Coding
`1.
`53. Kato describes the benefits of variable-length coding, such as
`
`Huffman coding, which it calls “inherently efficient.” Ex. 1002, Kato, 2:45. Kato
`
`then goes on to identify and address the same shortcomings with error propagation
`
`in variable length encoding purportedly addressed years later by the 482 patent:
`
`According to such variable-length encoding and related decoding, if
`an error occurs in the transmission of code words from an encoding
`side to a decoding side, the lengths of code words following the error
`can not be detected and also the boundaries between the code words
`following the error can not be detected. Thus, in the presence of an
`error, the code words following the error can not be accurately
`decoded. This problem is referred to as error propagation (see C.
`Yamamitsu, et al, “AN EXPERIMENTAL STUDY FOR A HOME-
`USE DIGITAL VTR”, IEEE Trans. CE-35, No. 3, Aug. 1989, pp 450-
`457).
`Id., 2:59-3:2 (emphasis added).
`
`54. Kato offers a number of solutions to address these shortcomings,
`
`many of which were later mirrored in the 482 patent. Among them, Kato teaches
`
`that individual code words can be split into first and second portions, wherein the
`
`first portion contains information about the second portion, such as its length, or
`
`17
`
`Page 18 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`the range of values in which it is located, which aids in preventing error
`
`propagation.
`
`55. Kato’s first embodiment is directed to the creation of coded data (Ci)
`
`to represent the difference (Si) between a given estimate (Pi) and the actual input
`
`data for which the estimate is made. Ci itself consists of two portions. The first
`
`portion (CJi) is entropy encoded from a category index (Ji). The second portion,
`
`(CEi), is generated from remainder data (Ei). See generally Ex. 1002, Kato, 7:1-
`
`59; 8:6-8; 9:38-10:4; and 17:32-34.
`
`56. First, an estimate Pi is generated for the input data (Di). Subtracting
`
`Pi from Di generates an estimation error Si, as previously described. Id., 6:60-64.
`
`57. Second, this estimation error Si is classified according to a
`
`classification table, as set forth, e.g., in TABLE 1:
`
`
`
`18
`
`Page 19 of 49
`
`
`
`Id., TABLE 1, 7:34-54.
`
`U.S. Patent No. 5,850,482
`
` Kato first describes determining a “category index (the category
`
`number) Ji denoting the range or the category where an estimation
`
`error Si [for input data Di] is present.” Id., 7:5-7.
`
` From this, divisor data (OUi) is calculated, which as shown in
`
`TABLE 1, “corresponds to the category index Ji in a one-to-one
`
`manner.” Id., 7:16-17.
`
` Then, remainder data Ei is calculated by dividing Di by OUi and
`
`determining the remainder. Id., 7:21-28.
`
` Finally, “the category index Ji and the remainder data Ei are
`
`encoded.” Id., 7:29-30.
`
`58. Thus, the category index Ji and remainder data Ei represent the
`
`estimation error Si for the original data Di. These Ji and Ei values are
`
`subsequently separately encoded to generate values CJi and CEi, which are
`
`combined to form a code word Ci, as shown in a portion of Kato’s Figure 1(a):
`
`19
`
`Page 20 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`CJi – “first portion”
`
`
`
`59. First, Kato teaches that Ji is encoded to generate a first portion of code
`
`word Ci – called CJi – by “encod[ing] the category index Ji into a Huffman code
`
`(an entropy coding technique). The sub encoding circuit 109 outputs the Huffman
`
`code in a bit serial format. The output signal from the sub encoding circuit 109 is
`
`the coded data CJi.” Id., 10:45-49.
`
`CEi – “second portion”
`
`60. Next, Kato teaches that Ei is encoded to generate a second portion of
`
`the code word Ci – called CEi – using “[t]he sub encoding circuit 110,” which
`
`“determines a bit number Mi of tile [sic] remainder data Ei in accordance with the
`
`category index Ji by reference to Table 1” and then “outputs Mi lower bits of the
`
`20
`
`Page 21 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`remainder data Ei in a bit serial format. The output signal from the sub encoding
`
`circuit is the coded data CEi.” Id., 10:65-11:2. And, Kato makes clear that CEi
`
`represents Ei: “The reason why only the Mi lower bits of the remainder data Ei are
`
`outputted is that the remainder data Ei can be represented by Mi bits.” Id., 11:2-4.
`
`61. Thus, for any given set of remainder data Ei associated with a given
`
`category index Ji, Table 1 sets a “remainder word length” for CEi – which Kato
`
`refers to as Mi – based solely on the value of Ji. For example, referring to
`
`TABLE 1 above, for an estimation error Si that generates a category index value Ji
`
`of 4, the “remainder word length” Mi for CEi coded from the remainder data Ei
`
`would be 3. Id., 14:51 (“Since Ji=4, Mi=3”).
`
`Ci – “code word”
`
`62. Kato goes on to state that once CJi and CEi are separately encoded,
`
`they are sent to a multiplexer, where they are combined into coded data Ci (id.,
`
`9:66-10:4), which is subsequently transmitted by an output terminal (108 in Fig.
`
`1(a)) (id., 11:9-11).
`
`2.
`
`Kato’s Fourth Embodiment
`Shows Unequal Error Protection (UEP)
`For First And Second Code Word Portions
`63. Kato’s fourth embodiment further describes a transmitter that arranges
`
`one or more code words Ci into a data transmission format that separates the code
`
`words into first portions Pi and second portions Ri. Id., 23:46-57; 24:40-45 (note
`
`21
`
`Page 22 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`that Pi in embodiment 1, supra, represented the predicted value, but in embodiment
`
`4 Pi refers to a first portion of a code word).
`
`64.
`
`In Fig. 6(a), and its accompanying discussion, Kato teaches using an
`
`encoding circuit 602 to encode input data using a variable-length code, and placing
`
`that code in a data store region within a data transmission format.” Id., 23:54-57.
`
`Kato further teaches providing an ECC encoder 603 that “adds an error correction
`
`code to the output data from the encoding circuit.” Id., 23:57-59.
`
`65. Kato further teaches that these first portions Pi and second portions Ri
`
`can be separately stored in different areas within a data store region, such as the
`
`example shown in Figure 7:
`
`
`
`Id., Fig. 7; see also id., 24:66-25:8.
`
`66. Alternatively, Kato teaches that instead of using a single data store
`
`region for the first code word portions Pi and the second code word portions Ri,
`
`these portions can each be stored in separate data store regions:
`
`22
`
`Page 23 of 49
`
`
`
`U.S. Patent No. 5,850,482
`
`According to a second example of the other arrangements of variable-
`length code words into a data store region, first portions Pi and second
`portions Ri are arranged in separate data store regions.
`Id., 32:35-39.
`
`67. Kato teaches and claims that the first portions Pi in one portion of the
`
`data store region (or in their own data store region) are preferably provided with a
`
`higher level of error protection than the second portions Ri. Id., 31:65-67 (“It is
`
`preferable that the data portions which are made higher in error-correction ability
`
`agree with the first portions Pi.”); see also id., claim 9, 35:18-21 (“an ability of
`
`error correction with respect to the first portions in the data store region is higher
`
`than an ability of error correction with respect to the second portions in the data
`
`store region.”); see generally id., 31:51-62.
`
`68. Additionally, Kato teaches that it is preferable to limit these first
`
`portions Pi to the smallest size which can be used to determine the length of the
`
`code word:
`
`It is preferable that a first portion Pi is limited to a smallest portion
`which enables the determination of the word length of a variable-
`length code word. This design minimizes the area occupied by first
`portions which can undergo error propagation, and thus minimizes the
`value “x” so that it is possible to withstand an error at a highest
`degree.
`Id., 32:14-20. An example of such a code word