US 7812828

I'll verify current docket/status information via search, while relying on the authoritative full-text provided for the patent's substance.

Let me pull the independent claim language and check for any 2026 CAFC activity.

Let me retrieve the independent claim text for claims 1, 10, and 16.

US Patent 7,812,828 — Verified Summary

I searched for the specific number 7,812,828 (not similar numbers) and cross-checked the authoritative full text supplied in the brief against live web results. Here is what is confirmed, plus an explicit flag on what I could not verify verbatim in this session.

---

Bibliographic data (confirmed)

Note on the inventor name: one OCR'd copy of the printed patent in a court exhibit renders the second inventor as "John G. Ellas." That is an OCR artifact; the authoritative spelling is John G. Elias, consistent with Google Patents, Justia, and the court filings.

Abstract (verbatim)

"Apparatus and methods are disclosed for simultaneously tracking multiple finger and palm contacts as hands approach, touch, and slide across a proximity-sensing, multi-touch surface. Identification and classification of intuitive hand configurations and motions enables unprecedented integration of typing, resting, pointing, scrolling, 3D manipulation, and handwriting into a versatile, ergonomic computer input device."

---

Plain-language overview of the independent claims

Important caveat: I verified claims 24, 29, 31, 32, 34, 35 verbatim (from the issued patent and from the Apple v. Samsung / ITC record). I did not retrieve the exact verbatim text of claims 1, 10, 16, and 23 in this session. For those I give the substance as best understood and flag the description as reconstructed rather than quoted. Claim numbering per the printed patent: 1–9 = method claims; 10–15 and 16–22 = device claims; 23 = computer-readable medium; 24–35 = "means for" device claims.

Claim 1 — method (independent). A method for processing touch input on a surface: (a) producing a proximity image representing a scan of a plurality of electrodes of a touch-sensitive surface, the image having pixels corresponding to those electrodes; (b) segmenting the proximity image into one or more pixel groups, each group representing a touch object on or near the surface; and (c) fitting an ellipse to at least one of those pixel groups. (Substance; wording not independently verified this session.) This claim term was the subject of extensive construction litigation — see the ITC discussion below, where the tribunal held that "fit an ellipse" requires actually fitting an ellipse to the data, not merely computing parameters that could describe one.

Claim 10 — touch-sensing device (independent). A touch-sensing device that produces a proximity image from a scanned electrode array, segments the image into pixel groups representing touch objects, and includes a contact tracking and identification module adapted to fit an ellipse to at least one pixel group. (Substance; dependent claim 20 confirms the module is adapted to compute eigenvalues and eigenvectors to fit the ellipse; claim 21 depends on claims 10–12 and 16–18.)

Claim 16 — touch-sensing device (independent). A device having the proximity-image / segmentation / ellipse-fitting structure with a contact tracking and identification module, where the change in an ellipse parameter (e.g., position, shape, size, orientation, eccentricity, major radius, minor radius) is used as a control input/control signal to an electronic or electromechanical device (as recited in dependent claims 17 and 19). (Substance; dependent claims 17–20 verified.)

Claim 23 — computer-readable medium (independent). A machine-readable medium carrying instructions executable to perform the method of any of claims 1–9.

Claim 24 — touch-sensing device (independent, means-plus-function). Verbatim: "A touch-sensing device comprising: means for producing a proximity image representing a scan of a plurality of electrodes of a touch-sensitive surface, the proximity image having a plurality of pixels corresponding to the touch-sensing electrodes; and means for segmenting the proximity image into one or more pixel groups, each pixel group representing a touch object on or near the touch-sensitive surface; and means for fitting an ellipse to at least one of the pixel groups."

Claim 31 — device (independent, means-plus-function). Verbatim: adds (i) means for tracking a path of one or more pixel groups through a plurality of time-sequenced proximity images; (ii) means for fitting an ellipse to at least one pixel group in a plurality of successive images; and (iii) means for tracking a change in one or more ellipse parameters through those time-sequenced images (claim 32: transmitting that change as a control signal; claim 34: fabricated on or integrated with a display device; claim 35: display = LCD or light-emitting polymer display).

Claim 29 (depends on 24) adds transmitting ellipse parameters as a control signal; claim 30 makes the device integral with the electronic/electromechanical device.

---

Litigation and docket status

2026 CAFC check (as requested): I found no 2026 Court of Appeals for the Federal Circuit docket activity naming 7,812,828. This is consistent with the record showing the patent expired fee-related on January 1, 2021 (adjusted expiration), which makes new infringement appeals unlikely. The only 2026 Federal Circuit items my searches surfaced involved other Apple patents (e.g., Apple Inc. v. Squires, No. 24-1864, on Fintiv discretionary denial; various PTAB appeals) — none referenced 7,812,828. I cannot assert a negative with certainty; treat this as "nothing found," not "confirmed none."

Historical enforcement (confirmed via the Google Patents family litigation links and court records):

• ITC Investigation 337-TA-750 (Apple v. Motorola) — '828 asserted.
• Apple Inc. v. International Trade Commission, No. 12-1338 (Fed. Cir. Aug. 7, 2013) — the Federal Circuit vacated the ITC's determination that Motorola did not infringe the '828 patent, and remanded. (The same opinion affirmed anticipation of claims 1–7 of the '607 patent.) Search result corroboration: Finnegan's Last Month at the Federal Circuit, September 2013.
• N.D. Cal. Case 5:11-cv-01846-LHK (Apple v. Samsung) — '828 was originally asserted and then, per case records, Apple dismissed all claims of the '828 patent during claim-streamlining in 2012.
• Wisconsin Western District, 3:10-cv-00661; other CAFC appeals listed for the family: 12-1105, 12-1338, 12-1506, 12-1600, 12-1606, 14-1335, 14-1368, 15-1029, 15-1857, 15-2088; Supreme Court No. 15-777 (Apple v. Samsung).

Claim-construction note of substance: In the ITC/related proceedings, the tribunal construed "fitting an ellipse" to require actually fitting an ellipse to the pixel-group data, rejecting a broader reading that merely calculating parameters that could define an ellipse (e.g., a covariance matrix that could equally describe a rectangle) suffices. The tribunal declined, however, to limit the term to the covariance-matrix methodology described in the specification, since the specification does not make that the exclusive technique. This construction materially narrows the '828 claims relative to its face value.

---

Uncertainty statement

• Title, assignee, inventors, filing/issue/priority dates, abstract, application number, and continuity data are highly reliable (multiple independent sources agree).
• The verbatim wording of claims 1, 10, 16, and 23 was not independently retrieved in this session; I have described their substance and flagged it as such rather than presenting it as quoted text. Claims 24–35 are verbatim-verified.
• The absence of 2026 CAFC activity is a negative finding from search and should not be treated as conclusive.

I'll research this patent's litigation history using the sources provided.

Let me dig into the specific case identifiers from the patent record.

Scope, method and caveats

I researched only US Patent No. 7,812,828 ("Ellipse Fitting for Multi-Touch Surfaces," Westerman & Elias, originally FingerWorks, assigned to Apple Inc. on 2007‑09‑20). I did not expand to similar identifiers. In particular, I discarded search hits for US 8,782,282 (a K.Mizra LLC network‑management patent asserted against Ciena — a completely different patent, including IPR2025‑01362 and Unified's ex parte reexam 90/019,963). Those results are not about the patent you asked about.

One important limitation: the litigation list on the patent's Google Patents page is presented under "Family has litigation," which pools dockets across the entire FingerWorks family (priority 1998‑01‑26, application 09/236,513). Some dockets listed there may involve sibling patents ('607, '430, '381, etc.) rather than the '828 patent itself. I have therefore separated verified '828‑specific cases from family‑level docket entries I could not confirm are '828‑specific.

---

A. Litigation verified to involve the '828 patent

Notes on the ITC case and appeal

• The ITC construed "mathematically fit[ting] an ellipse" as requiring a mathematical process whereby "an ellipse is actually fitted to the data," and on that basis found non‑infringement (J.A. 58–70).
• The Federal Circuit's 2013 decision revived the '828 issue for remand. In May 2014 Apple and Motorola/Google agreed to dismiss all patent litigation between them, which (per press reporting) ended the '828 dispute without a final merits win for either side. I have medium confidence on the precise mechanics of that global dismissal.

---

B. Family-level docket entries shown on the patent record — NOT confirmed as '828‑specific

The Google Patents page lists the following additional dockets under its litigation section. I could not independently verify that US 7,812,828 (as opposed to a sibling family member) was the asserted patent in each. Treat these as leads, not ground truth:

• CAFC 15‑1029, 15‑2088, 15‑1857, 12‑1606, 12‑1600, 12‑1338, 12‑1506, 14‑1335, 14‑1368
• U.S. Supreme Court 15‑777 — this is Samsung Electronics Co. v. Apple Inc., which concerned design‑patent damages (art. of manufacture/"total profit"), not the '828 patent. It appears here only because of the Apple v. Samsung litigation overlay.
• The patent record's "First worldwide family litigation" entry (Darts‑ip family 26753441).

On the Federal Circuit appeal number for the ITC case: the patent page lists 12‑1105, and the ITC appeal was filed in April 2012, which is consistent with a 2012 docket number. I have medium confidence in that mapping; I did not retrieve the CAFC docket itself.

---

C. Express negatives

• Apple v. HTC: I found no evidence that US 7,812,828 was asserted against HTC (in the D. Del. actions or ITC 337‑TA‑710). Apple's early HTC assertions centered on other patents.
• Motorola's retaliatory actions against Apple (W.D. Wis. 3:10‑cv‑00642 and ITC 337‑TA‑745, both October 2010) asserted Motorola's own patents; I found no indication '828 was at issue in them.
• I found no IPR, PGR, CBM, or ex parte reexamination directed at US 7,812,828 in the materials reviewed. (The IPR/reexam hits in my searches were for the unrelated 8,782,282 patent.)

---

D. Patent status (for context, from the patent record)

• Current assignee: Apple Inc.; original assignee Apple Inc.; assignors FingerWorks, Inc. (recorded 2007‑09‑20).
• Status: Expired – Fee Related; Google Patents shows an adjusted expiration of 2021‑01‑01. Note a discrepancy: Apple's own 2010 ITC complaint stated the '828 patent expires January 25, 2019. I am reporting both literally rather than reconciling them.

---

Bottom line

US 7,812,828 was asserted in litigation in essentially one campaign: Apple's October 2010 multi‑touch enforcement against Motorola, pursued in parallel in the W.D. Wis. (3:10‑cv‑00661) and at the ITC (337‑TA‑750), with the ITC branch appealed to the Federal Circuit (725 F.3d 1356 (Fed. Cir. 2013)). The '828 patent also appears in the Apple v. Samsung N.D. Cal. 5:11‑cv‑01846 matter, but only at the claim‑construction stage and it was not tried to verdict. No case produced a final, upheld merits judgment of infringement on '828; the ITC found non‑infringement (reversed and remanded on claim construction in 2013), and the disputes were ultimately dismissed/settled.

If you want, I can attempt to pull the actual Unified Patents litigation portal page for 7,812,282's counterpart family or the PACER dockets for 3:10‑cv‑00661 to confirm the dismissal dates — but I did not verify those docket-level details here, so I've flagged them rather than asserted them.

I'll verify the structured block's "no AIA trials" finding with targeted searches before finalizing.

The first pass returned only noise (unrelated patents 8,782,282 and 8,281,425). Let me run tighter queries to make sure I'm not missing a proceeding under a variant citation.

Searches consistently show this patent's challenges were in the ITC and district courts, not the PTAB. Let me run two final confirmatory queries, including for any ex parte reexam or CBM record.

Proceedings overview

Zero AIA trial proceedings — the USPTO Open Data Portal structured block returns no IPR, PGR, or CBM for US 7,812,828, and my independent searches converged on the same null result: 0 active / 0 claims invalidated / 0 claims sustained / 0 settled on the PTAB record / 0 institution denials.

Because that count is zero, the per-proceeding template below has nothing to fill in — so I adapt it. There is one thing a defendant must read before relying on "no PTAB activity = untested patent": the validity and claim-construction fight over the '828 patent actually happened, just in the ITC and the Federal Circuit, not at the Board. The bottom-line defensive posture is therefore neither of the two poles in the prompt. It is: no claims have been canceled, so you cannot kill an infringement theory by pointing at an FWD — but the only court to construe "mathematically fitting an ellipse" read it narrowly, against Apple, and Apple did not get a clean win on this patent anywhere.

---

Proceedings on file (USPTO ODP): none

The canonical structured list is empty. No IPR, no PGR, no CBM. Consistent with that:

• No § 315(e)(2) estoppel attaches to any party, because no AIA trial was ever instituted.
• No PTAB Final Written Decision exists to cite, quote, or link.
• No Federal Circuit appeal from a PTAB FWD exists (the CAFC dockets in the family — 12-1105, 12-1338, 12-1506, 12-1600, 12-1606, 14-1335, 14-1368, 15-1029, 15-1857, 15-2088 — are appeals from the ITC and district court, not from the Board).

Verification path and caveat: the ODP "no proceedings" result is the canonical source, and my web searches (querying "7812828" IPR OR CBM OR "post-grant review", "IPR2013"/"IPR2014"/"CBM2013" "7,812,828", and Apple/Motorola-specific combinations) surfaced only unrelated patents — 8,782,282 (Ciena v. K.Mizra, IPR2025-01362) and 8,281,425 (IPR2014-00190). Those are not this patent; do not confuse the digits. This is a negative finding, and negative findings are weaker than positive ones — but it is corroborated by the procedural history below.

Why the '828 patent never drew an IPR — the explanatory analogue

Not a PTAB proceeding. Included because a defendant needs it, and because it is the closest thing to a validity adjudication this patent has.

Apple Inc. v. International Trade Commission, No. 2012-1338 (Fed. Cir. Aug. 7, 2013) — appeal from ITC Investigation 337-TA-750 (Apple v. Motorola).

• Claim construed: Claim 1, verbatim per the Federal Circuit opinion: "A method of processing input from a touch-sensitive surface, the method comprising: receiving at least one proximity image representing a scan of a plurality of electrodes of the touch-sensitive surface; segmenting each proximity image into one or more pixel groups that indicate significant proximity, each pixel group representing proximity of a distinguishable hand part or other touch object on or near the touch-sensitive surface; and mathematically fitting an ellipse to at least one of the pixel groups."
  • Correction to the prior section of this analysis: the earlier summary flagged the wording of claim 1 as "not independently verified." The Federal Circuit opinion supplies it verbatim, quoted above. The one-shot "receiving / segmenting / mathematically fitting" three-step structure is confirmed. Claim 10 of the '828 patent was also identified in the same opinion as requiring a "pixilated image."
• Disposition on '828: The ITC found Motorola did not infringe, construing "mathematically fitting an ellipse" to require "a mathematical process" whereby "an ellipse is actually fitted to the data." The Federal Circuit vacated and remanded on the '828 and '607 patents, holding the ITC's analysis defective (including a hindsight-obviousness concern on '607); it affirmed the ITC on seven other '607 claims. (FindLaw opinion; PTAB E2E negative check: https://ptacts.uspto.gov/ptacts/; CourtListener — search Apple v. International Trade Commission, 12-1338.)
• Flag: the prior section also stated the ITC "declined to limit the term to the covariance-matrix methodology of the specification." I did not re-verify that sub-point this session. Treat the "actually fitted to the data" construction as verified; treat the covariance-matrix carve-out as carried over and unconfirmed.
• Defensive value today: the only judicial construction of the '828 claim term is the narrow one — "actually fitted to the data." That is a noninfringement theory, not an invalidity theory, and it survives the patent's expiration because it does not depend on the patent being valid. Apple also dismissed all claims of the '828 patent in Apple v. Samsung, N.D. Cal. 5:11-cv-01846, during 2012 claim-streamlining — i.e., Apple repeatedly chose not to carry this patent to judgment.

---

Strategic summary

Canceled vs. sustained vs. untested. All 35 claims of US 7,812,828 are UNTESTED at the PTAB. Zero are canceled; zero have been held patentable in an AIA trial. The claims were adjudicated only for infringement/claim construction (ITC, remanded) and were dropped from the one district court case where they might have reached a verdict. Adding the bibliographic context from the prior section: the patent expired fee-related on 2021-01-01 (707 days of § 154(b) term adjustment). Whatever posture a demand letter claims, the enforceable tail on this patent is gone — there is no ongoing royalty base and no injunction to fear from an assertion today.

Estoppel landscape. This is the cleanest part of the picture for a defendant: with no IPR ever instituted, no § 315(e)(2) estoppel binds anyone — not Motorola, not Samsung, not any prior defendant, and therefore no prior-art ground is foreclosed to you by the Board. Any printed publication or patent ground you can assemble is available to you. Offsetting that benefit: you get no free kill. There is no FWD to borrow, no canceled claim to knock out a count, and no PTAB record to cite as res judicata-adjacent persuasion. If a plaintiff asserts claim 1 today, you must win on your own record — and your strongest lever is inherited from the ITC: the narrow construction of "mathematically fitting an ellipse."

Pattern signals. No serial-petitioner pattern (there are no petitioners). No defensive aggregator (Unified Patents, RPX) appears anywhere in the '828 chain — the Google Patents family litigation links show only Apple-versus-Motorola/Samsung/HTC litigation and the ITC/CAFC/Supreme Court appeals (including Supreme Court No. 15-777). Apple did not pursue PTAB appeals on this patent. Timing explains the gap: the '828 patent's assertion window (2010–2012) largely predates the mature IPR regime (first IPRs filed 2012, effective 2012-09-16), and by the time IPRs became routine, the patent was in its twilight and Apple had dropped it from its flagship case.

Why an IPR filed today would be an uphill and probably futile exercise. Even ignoring expiration, three barriers stack up, and I flag these as policy context from search results, not holdings binding on this patent: (1) the § 315(b) one-year bar for anyone already served; (2) the PTO's 2025 discretionary-denial framework and its "settled expectations" doctrine, which — per Unified Patents' 2026 amicus brief in Sup. Ct. No. 25-1230 — has been invoked "to deny hundreds of IPR petitions" and disproportionately immunizes older patents; and (3) § 314(a)/§ 325(d) discretion against re-reviewing art already before the Office. A 1998-priority patent that expired in 2021 sits squarely in the category the 2025 guidance shields.

---

Recommended next steps

1. Do not represent that any claim of the '828 patent has been canceled. It hasn't. If opposing counsel or a demand letter implies PTAB invalidation of this patent, that is false, and the ODP record (PTAB E2E) disproves it in one query.
2. Lead with expiration, then construction. Check the expiration first — 2021-01-01, adjusted, expired fee-related — because it moots damages and injunctive relief. Then raise the ITC construction from Apple v. ITC, No. 2012-1338: "mathematically fitting an ellipse" requires a mathematical process in which "an ellipse is actually fitted to the data." Any accused product that derives contact parameters without actually fitting an ellipse (e.g., covariance/eigenvalue parameterization alone) has a strong noninfringement position on claim 1. Preserve that argument; it is the single most valuable item in this file.
3. If you nonetheless need an invalidity position, you are unconstrained by estoppel — no § 315(e)(2) bar applies. Mine the ITC record from 337-TA-750 and the remanded proceedings for the art Motorola actually ran, plus the references listed in the '828 prosecution history (the patent cites a dense field of capacitive-touch art, including Gillespie and Philipp). Because this would be a district-court invalidity case, remember claims 1–9 and 23–35 carry different structures, and the means-plus-function claims (24–35) require corresponding-structure analysis under § 112(f).
4. No trial-stage milestones to track. With zero instituted proceedings, there is no institution deadline, no oral hearing, and no statutory one-year FWD clock running. The absence of PTAB activity is itself the signal here: a patent asserted in three forums over a decade, against the largest Android OEMs, with no IPR ever filed against it, is a strong indicator that practitioners did not view the prior art as IPR-grade — or that the patent's narrow construction made invalidity work unnecessary. Do not read it as the patent being hardened by PTAB survival. It never entered the arena.

Confidence statement: The "no PTAB proceedings" conclusion is high-confidence for the ODP record and well-corroborated by search, but it rests on a negative finding. The claim 1 text and the "actually fitted to the data" construction are quoted from the Federal Circuit opinion and are reliable. The covariance-matrix carve-out carried over from the prior section is unverified this session. I did not find, and cannot rule out, an early, low-visibility petition that never produced a public institution decision.

I'll research the assignment chain for US 7,812,828 through multiple sources.

Let me look for the specific USPTO assignment reel/frame data.

I found a key reel reference. Let me pin down the specific reel/frame for this patent's application.

I hit the session's tool-call limit while retrieving the object-level Assignment Center record, so I'm reporting what the retrieved sources establish and explicitly flagging the one field I could not pull directly. No tool calls remain.

Sourcing note: The Assignment Center UI itself (https://assignmentcenter.uspto.gov/) was not directly queryable in this session. The chain-of-title below is reconstructed from a *USPTO assignment/chain-of-title statement filed in Apple Inc. v. [Samsung Electronics Co.](/litigations/by-defendant/Samsung%20Electronics%20Co.), N.D. Cal. 5:11-cv-01846 (Doc. 562-1)* — which recites USPTO reel/frame numbers for this exact application (11/677,958) — cross-checked against Google Patents legal events, the INPADOC/Espacenet register for the sibling patent US 8,576,177, and the court record. Where a reel/frame is confirmed I say so; where I could not confirm it I say that too.

---

Inventors

• Both inventors were the original owners of record of the '828 family; the assignment statement in the Samsung case recites the very first link as "From: Wayne Westerman; John G. Elias … Reel 009735, Frame 0932."
• No fire-sale/abandonment pattern. The inventors did not depart within 12 months of filing — the opposite. Wikipedia and the Apple opening brief in Apple v. ITC (No. 12-1338) both record that Apple acquired the company's IP along with Elias and Westerman, and that the founders "continued to file and process patents for their work through late 2007" and "as of August 2008 they still filed patents for Apple, Inc." That is an acqui-hire, not a portfolio dump.
• University-origin wrinkle: the chain includes the University of Delaware and its tech-transfer arm (UD Technology Corporation). This is a technology-transfer chain, not an NPE-style chain.

Original assignee

• Entity named on the issued patent: Apple Inc. (Cupertino, CA). Google Patents lists Apple Inc. as both "Original Assignee" and "Current Assignee"; the reassignment event is dated 2007-09-20, assignor "FINGERWORKS, INC."
• Did it ship a product embodying the claims? Yes, unambiguously. Apple shipped the iPhone (2007) and iPad (2010) multitouch stack; the '828 disclosure is the ellipse-fitting front end for exactly that. Apple's own appellate brief describes the patent as "the '828 patent — covering FingerWorks' ellipse-fitting multi-touch process."
• Predecessor's business (FingerWorks, Inc.): gesture-recognition hardware — the TouchStream multitouch keyboard, iGesture Pad/NumPad, and MacNTouch surface; Newark, Delaware; founded 1998; ceased shipping and went defunct in June 2005 after Apple acquired its assets in early 2005 (Apple paid ~$13.5M per the damages record in the Samsung case).
• Current status of assignee: Apple Inc. — operating, investment-grade public company. FingerWorks, Inc. — dissolved/absorbed, no longer a going concern (Company's products "remained a high-end niche," then it stopped shipping).

Assignment timeline

Chronological, per the recorded/reported links:

• c. 1998–1999 (executed) / recorded c. 1998–1999 — Reel 009735/0932

  • Conveyance: Assignment
  • Assignor: Wayne Westerman; John G. Elias
  • Assignee: not legible in the excerpt I retrieved — the OCR of the Samsung chain-of-title statement truncates the "To:" field. Since this is link 1 of a chain that terminates at Apple, it is most likely FingerWorks, Inc., but I could not verify this and will not assert it as fact.
  • Correspondent: not retrieved. Unknown. (Flagged as a gap, not as a finding.)
  • Context: inventor-to-company assignment at company formation.

• c. 1999–2001 (executed) / recorded c. 1999–2001 — Reel 016521/0486

  • Conveyance: Assignment
  • Assignor: University of Delaware
  • Assignee: UD Technology Corporation (university tech-transfer arm; assignee name inferred from the parallel UD→FingerWorks link below and the UD license agreement dated 7/19/1999 in the Apple production)
  • Correspondent: not retrieved.
  • Context: university technology-transfer internal step (UD separating its rights into its commercialization entity).

• c. 2000–2001 (executed) / recorded c. 2000–2001 — Reel 016536/0701

  • Conveyance: Assignment
  • Assignor: UD Technology Corporation
  • Assignee: Fingerworks, Inc.
  • Correspondent: not retrieved.
  • Context: university technology-transfer — UD's rights consolidated into FingerWorks; corroborated by a July 19, 1999 "License agreement between the University of Delaware, and FingerWorks of Delaware" in Apple's production.

• 2007-08-31 (executed/effective) / recorded 2007-09-20 — Reel/frame NOT CONFIRMED for '828 (see note)

  • Conveyance: Assignment of Assignors Interest (portfolio acquisition of the FingerWorks patents)
  • Assignor: FingerWorks, Inc.
  • Assignee: Apple Inc.
  • Correspondent: not retrieved from the Assignment Center. Note: the prosecution correspondence address of record for application 11/677,958 was Gregory S. Weaver (Reg. No. 53,751), Morrison & Foerster LLP, 555 West Fifth Street, Los Angeles, CA 90013-1024 (attorney docket 10684-25086.04) — an operating-company firm doing Apple's prosecution, not an asserter-side filer. Do not conflate the prosecution correspondent with the assignment-recorded correspondent; I could not confirm the latter.
  • Context: bona fide corporate acquisition — Apple's purchase of FingerWorks' IP (deal announced/closed 2005; formal USPTO recordation of the patent batch executed 2007-08-31).
  • Reel/frame note (flagged): the sibling FingerWorks→Apple record for US 8,576,177 is Reel 020063/0971, executed 2007-08-31, recorded 2007-11-02 (confirmed via INPADOC/Espacenet). Apple recorded the FingerWorks acquisition in batches — third-party litigation exhibits list FingerWorks, Inc. → Apple records with the same 2007-08-31 execution date but recording dates spread across 2007-09-10, 2007-09-19, 2007-09-21, 2007-11-02 and 2007-11-14. The '828 record (recorded 2007-09-20 per Google Patents) therefore sits in a different reel from 020063/0971, in approximately the 0199xx range. I did not retrieve the exact number and will not guess it.

No post-Apple assignments found. I found no record of the '828 patent leaving Apple — no security agreement, no merger, no license recordation, no transfer to any LLC or to a defensive aggregator.

Timeline diagram

timeline
    title Ownership of US 7812828
    1998 : Elias and Westerman found FingerWorks Inc
         : Provisional application filed
    1999 : First non-provisional application filed
    2000 : University of Delaware rights assigned
    2005 : Apple acquires FingerWorks assets
    2007 : Continuation application filed Feb 22
         : FingerWorks assigns patent rights to Apple
    2010 : Patent issues Oct 12
         : Apple files ITC complaint against Motorola
    2013 : Federal Circuit vacates ITC ruling
    2021 : Patent expires Jan 1

NPE / troll-pattern signals

1. Shell-entity transfer — NOT PRESENT. The chain runs operating-company → operating-company (FingerWorks, Inc. → Apple Inc., effective 2007-08-31). There is no "IP / Holdings / Licensing / Ventures" entity anywhere in the chain. FingerWorks shipped real hardware (TouchStream, iGesture Pad); Apple ships the accused-into-existence products.

2. Known asserter in the chain — NOT PRESENT. Neither FingerWorks nor Apple appears on the Acacia / Marathon / IV / IPNav / Wi-LAN / Conversant / Vringo / Round Rock / MPHJ lists. Apple is a practicing entity that asserts its own patents in its own name.

3. Repeat correspondent across the chain — NOT PRESENT / UNCLEAR (insufficient data). There is only one post-issuance link (FingerWorks→Apple), so recurrence cannot even be tested. The only attorney I could positively identify is Gregory S. Weaver, Reg. No. 53,751, Morrison & Foerster LLP (prosecution of 11/677,958), which is standard operating-company prosecution work, not asserter-side recordation. A single appearance is not a finding under your own rule, and I did not retrieve the assignment-recorded correspondent. Marked unclear because the correspondent field is blank in my data, not because anything suspicious appeared.

4. Cascading transfers — NOT PRESENT. Exactly one substantive transfer of the patent (2007), plus the pre-existing formation/tech-transfer links. The 2007-09-10 → 2007-11-14 recording dates are a single acquisition recorded in batches, not chained LLC hops — all share the same 2007-08-31 execution date and the same assignor/assignee pair.

5. Pre-litigation transfer — NOT PRESENT. Assignment effective 2007-08-31; first assertion naming '828 was Apple's ITC complaint filed 2010-10-29 (Investigation 337-TA-750). That is ~38 months, far outside the 6-month window. If anything the causality runs the other way: the patent sat in Apple's portfolio for three years before Apple invoked it.

6. Bankruptcy fire-sale — NOT PRESENT. FingerWorks was acquired (Apple, early 2005, ~$13.5M), never in Chapter 7/11. No bankruptcy docket appears in the family's litigation links for this patent.

7. Privateering — NOT PRESENT. Apple asserted the '828 itself, against its own competitors (Motorola at the ITC; HTC; originally Samsung in 5:11-cv-01846 until Apple dropped all '828 claims in 2012 claim-streamlining). There is no separate NPE asserting on Apple's behalf, and no SEC 10-K/8-K disclosure of a patent-transfer arrangement to an asserter.

8. Defensive aggregator — NOT PRESENT. The chain terminates at Apple Inc., an active litigant that vacated and remanded (Fed. Cir. No. 12-1338, Aug. 7, 2013) the ITC's no-infringement finding on '828. The patent was never neutralized by RPX, AST, LOT, Unified or OIN.

Verdict

Operating-company assertion.

The chain is a textbook university spin-out → startup → acquirer sequence — University of Delaware (Reel 016521/0486; Reel 016536/0701) and inventors Westerman/Elias (Reel 009735/0932) into FingerWorks, Inc., then a single bona fide acquisition of the FingerWorks IP by Apple Inc. executed 2007-08-31 (recorded for '828 on 2007-09-20; sibling record Reel 020063/0971 for US 8,576,177, same execution date, recorded 2007-11-02). The current assignee is a practicing, investment-grade manufacturer that itself asserted the patent in its own name against Motorola (ITC 337-TA-750; Fed. Cir. 12-1338) — the definition of operating-company assertion, not NPE activity. Zero of the eight NPE signals is present; signals 3 and 4 are the only ones I could not fully close, and only because the Assignment Center's correspondent and '828-specific reel/frame fields were not retrievable in this session.

Verify at: https://assignmentcenter.uspto.gov/ (search "7812828") — also indexed at https://assignment.uspto.gov/patent/index.html.

Explicit gaps (not negative findings):

• The exact reel/frame of the FingerWorks→Apple record for '828 and the assignment-recorded correspondent were not retrieved. I confirmed the sibling reel 020063/0971 and the batch pattern inference; I did not obtain the '828 number itself.
• The assignee on the first link (Reel 009735/0932) is illegible in the source I used and is inferred from chain position only.
• Any 2026 maintenance/ownership activity is unverified; the register indicates the patent expired fee-related on 2021-01-01 (adjusted expiration), which is consistent with no further transfers.

I'll search for the specific patent number and its cited references.

Let me retrieve the full "References Cited" list from the patent's front page and related databases.

I have partial citation lists. Let me retrieve the complete "References Cited" lists from the patent's front page.

I've hit the session's search-step limit. Below is the analysis based on what I was able to verify, with explicit flagging of what I could not retrieve. I did not fabricate any part of the citation list.

---

Prior-Art Analysis — US 7,812,828 B2 ("Ellipse fitting for multi-touch surfaces")

0. Sourcing, and an honest statement of limits

I searched for the literal string 7812828 and did not broaden to similar numbers. What I could reach:

• Google Patents full text of US7812828B2 (authoritative copy supplied in the brief).
• uspto.report/patent/grant/7812828 — a USPTO-data mirror that returns the front-page "References Cited" table.
• Court exhibits reproducing the printed patent's front page (Apple v. Samsung, N.D. Cal. 5:11-cv-01846; Justia/CourtListener PDFs), which show the "(56) References Cited" block.

What I could NOT do — flagging rather than papering over:

1. I could not run a live USPTO PatentCenter/PAIR query in this session.
2. I recovered the citation list in fragments, not as a single ordered verbatim block. There is a gap between US 5,241,308 (Aug. 1993) and US 5,675,361 (Oct. 1997), and the recovered list appears to stop at US 6,002,808 (Dec. 1999). Do not treat my list as the complete "References Cited" section.
3. For most entries I recovered only number / issue date / inventor surname. I did not retrieve per-reference titles or claim text this session, so I mark those as "bibliographic only" rather than inventing descriptions.

Identifier-integrity flag (per the literal-interpretation rule): one search returned a different patent bearing the number 7812828 — Japanese Patent No. 7812828 (特許第7812828号), an "electrochemical reaction single cell / cell stack," application JP 2023-171108, published 2025-04-14. That is not the US patent and is unrelated. I did not merge it in. Separately, a minor date inconsistency exists across sources: Google Patents gives the earliest priority as 1998-01-26, while Unified Patents and golden.com render it 1998-01-25. I flag this rather than reconcile it.

---

1. Critical framing you need before the tables: cited ≠ anticipating

Three points that materially change how the requested analysis should be read:

(a) The "(56) References Cited" block is a disclosure list, not a novelty holding. It contains (i) art the examiner used to reject, (ii) art cited for background/context, and (iii) art the applicant disclosed under the duty of candor. Most of the ~100 U.S. patent documents on the '828 front page were background citations. Treating each as "potentially anticipating" overstates the list.

(b) §102 requires date qualification against the correct critical date. The '828 claims benefit to 09/236,513 (filed Jan 25, 1999), which claims benefit of provisional 60/072,509 (Jan 26, 1998). Working dates:

• Pre-AIA §102(a)/(b) window: references published before ~Jan 26, 1997 (grace date) — the large 1967–1996 block.
• §102(e) window: U.S. patents/printed publications filed before Jan 26, 1998 but issuing later — this is where most of the 1997–1999 entries in the list fall, but only if their filing dates predate the critical date. I could not verify individual filing dates this session, so each is a candidate, not a conclusion.
• Post-1998 issue dates do not by themselves make a reference prior art.

(c) Ellipse fitting is the point of novelty, and none of the pre-1998 art I recovered appears to teach it. The claim terms that matter (as construed in the ITC/Apple v. Motorola line, per the earlier section) are proximity image → segment into pixel groups → actually fit an ellipse and, in claims 17–19/31–32, use a change in an ellipse parameter as a control signal. A reference that computes a centroid of a single object, or a covariance/hashing parameter set, does not anticipate those limitations — that was precisely the construction dispute.

---

2. Table A — U.S. Patent Documents cited (recovered portion)

Format: number / issue date / inventor, then what it is, then §102 posture. Entries marked ⓑ are bibliographic-only (I have number/date/name, not verified subject matter).

Block I — 1967–1992 (potential §102(a)/(b) art, pre-1997)

Gap: citations between US 5,241,308 (Aug. 1993) and US 5,675,361 (Oct. 1997) were not recovered. This is a meaningful gap — it is where the Synaptics/Gerpheide/Boie/Gillespie/Logan art most likely sits (see §3).

Block II — Oct. 1997 – Dec. 1999 (the overlapping-art block; §102(e) candidates)

This block is at/around the critical date and is where genuine §102 risk concentrates. Issue dates run after Jan 26, 1998, so their status depends entirely on filing dates I could not verify.

---

3. Table B — Foreign patent documents (recovered)

I could not recover the remainder of the foreign-document block (e.g., JP/EP/WO entries) this session — do not assume CA 1243096 is the sole foreign citation other than what the recovered fragment shows.

---

4. Table C — Other publications ("Other Publications" block)

These are the references I recovered with enough content to characterize. This block is where the non-patent prior art lives, and it is genuinely probative for the algorithmic claim limitations.

---

5. Bottom line: what is "most relevant," ranked

Lead §102 candidates (date-dependent, all requiring filing-date verification):

1. US 5,825,352 — Bisset et al. (Synaptics), Oct. 20, 1998. The closest thing on the face of the patent to a multi-finger contact-sensing anticipation. Targets the "proximity image → segment into groups, each group representing a touch object" and multi-contact-tracking limitations (claims 1, 24, 31 as construed).
2. US 5,920,309 — Bisset et al. (Synaptics), Jul. 20, 1999. Companion to '352; same targeting.
3. US 5,730,165 — Philipp, Mar. 10, 1998. Charge-transfer self-capacitance sensing — targets the sensor means of claims 24–30.
4. US 4,746,770 / 4,968,879 — McAvinney. Multi-object sensing; targets the "segment into groups each representing a touch object" limitation.
5. US 5,880,411 — Gillespie et al. (Synaptics), Mar. 9, 1999. Interpolated centroid position from row/column self-capacitance; targets the proximity-image/centroid aspects.

Lead §103 references (combination): Bisset '352 + Gerpheide '457 (5,767,457) + Gillespie '411 + McAvinney '770/'879 + Chang 1993 (nearest-neighbor tracking) + Hillier & Lieberman (optimization).

Best non-patent §102(b) candidates (pre-1997): Chang et al. 1993 (nearest-neighbor searching for the path-matching step); Davis et al. 1994 (3-D hand motion); Buxton et al. 1985.

Explicit conclusion on the requested §102 mapping: On the recovered evidence, no cited reference anticipates the ellipse-fitting limitations themselves — i.e., the "fit an ellipse to at least one of the pixel groups" element of claim 1/24 and the "tracking a change in one or more ellipse parameters" element of claims 31–32. The citing references attack the sensing means and the segmentation/multi-contact tracking, not the ellipse fit. That is consistent with the earlier-reported construction in the ITC/Apple v. Motorola line, where "fitting an ellipse" was held to require actually fitting an ellipse, and with the fact that the '828 survived that proceeding (the Federal Circuit vacated non-infringement in Nos. 12-1338).

---

6. What remains unverified — stated plainly

• Incomplete citation list: the mid-1993→1997 gap (between US 5,241,308 and US 5,675,361), the tail after US 6,002,808, the balance of the FOREIGN block, and the remainder of the "Other Publications" block (I recovered the entries above; the justia fragment shows the list continues past "Jacob et al., Integrality and Separability of Input Devices, ACM TOCHI").
• No verified filing dates for the 1997–1999 references; therefore the §102(e) designation for each is a candidate, not a finding.
• Two specification-discussed references (Boie US 5,463,388; Gillespie US 5,543,591; Logan US 5,327,161; Gillick US 5,530,455; the "Lee" 32×64 multi-touch array) are expressly analyzed in the '828 written description. They almost certainly appear in the 1993–1997 gap of the cited list, but I did not confirm their presence on the front page this session — so I am reporting them as specification-cited art rather than as front-page citations. Given the specification's own characterizations, Boie et al. (capacitive sensing array usable for keyboard or mouse) and Gillespie et al. (per-electrode self-capacitance with interpolated position) are the two most material sensing-side references the applicant itself distinguished over.
• 2026 activity: consistent with the earlier section, nothing surfaced tying 7,812,828 to 2026 Federal Circuit activity; the patent's adjusted expiration is Jan. 1, 2021.

I'll gather the prior-art references cited in this patent's record before building the § 103 combinations.

Now let me pull the prosecution history and the specific prior-art references that were applied during examination.

Let me pull the specific § 103 rejections from the file history and check the invalidity art Motorola ran at the ITC.

Obviousness Analysis — U.S. Patent 7,812,828 under 35 U.S.C. § 103

Analyst note on inputs: I was directed to "the Prior Art section of this page." The Google Patents record for US 7,812,828 exposes a prior-art keyword set (touch, hand, finger, proximity, fingers) and a References Cited list, and the Justia/Espacenet records reproduce that list plus an "Other Publications" section. I have therefore built this analysis on (a) the references cited of record in the '828 patent, and (b) the actual § 102/§ 103 rejections from the '828 file history (application 11/677,958), which I retrieved from the file-wrapper exhibits filed in Apple Inc. v. [Samsung Electronics Co.](/litigations/by-defendant/Samsung%20Electronics%20Co.), N.D. Cal. No. 5:11-cv-01846 (Dkt. 562 and 542, exhibits 14 and 15). Where I quote the Examiner I am quoting those exhibits.

---

0. Two threshold flags before the analysis

Flag A — Contradiction with the previously generated section (must be resolved in favor of the search results).

The earlier section asserts, and builds its "recommended next steps" on, the proposition that:

"the only court to construe 'mathematically fitting an ellipse' read it narrowly, against Apple… the ITC held that 'fit an ellipse' requires actually fitting an ellipse… Preserve that argument; it is the single most valuable item in this file."

The live results contradict this as a current statement of law. In Apple Inc. v. International Trade Commission, 725 F.3d 1356 (Fed. Cir. 2013):

• The ITC had construed "mathematically fitting an ellipse" to require "a mathematical process" whereby "an ellipse is actually fitted to the data," and on that basis found Motorola did not infringe.
• The Federal Circuit overturned that construction. Reporting on the decision states the court "first overturned the ITC's construction of the term 'mathematically fitting an ellipse' and instead agreed with Apple's construction," and that the noninfringement judgment was vacated and remanded "for consideration under the correct construction." Apple's own brief frames the issue as whether the ITC erred by "superimposing on the claim term 'mathematically fitting an ellipse'… the anachronistic requirement that the software 'actually' fit an ellipse."

Practical consequence for a § 103 analysis: the claim is broader than the earlier section assumed. A broader "mathematically fitting" construction increases the prior-art coverage of claim 1 and decreases the noninfringement leverage. I flag this because the earlier section made the opposite assertion, and because it changes the obviousness calculus materially. (Note: the two earlier flagged sub-issues — the "covariance-matrix carve-out" — remain unverified in this session; I did not re-verify it here either.)

Flag B — OCR variance in the reference list. The two court exhibits and the Justia list disagree on several digits of the cited patents (e.g., "5,189,400" vs. "5,189,403" for Franz; "5,281,906" vs. "5,281,966" for Walsh; "4,733,222" vs. "4,733,422"). Per the standing rule, I do not auto-correct these. Where a digit string is load-bearing for the analysis below, I say so.

---

1. Legal framework

• Statute: pre-AIA 35 U.S.C. § 103(a). The '828 patent carries a 1998-01-26 priority date and its specification was filed pre-AIA. The file history confirms the Examiner applied the pre-AIA framework (the office actions quote pre-AIA § 103(a) and cite Graham).
• Test: Graham v. John Deere Co., 383 U.S. 1 (1966) — (i) scope and content of the prior art; (ii) differences between the claims and the prior art; (iii) level of ordinary skill; (iv) objective indicia. KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007) — the combination need not be taught expressly; a "predictable result," a "known need," or "common sense" suffices, and the TSM test is not to be applied rigidly.
• Claim construction drives the analysis. Under the Fed Cir's rejection of the ITC's narrow construction, "mathematically fitting an ellipse" is satisfied by computing numerical parameters that mathematically define an ellipse — not only by an "actual" conic fit. I note, however, that the specification's own ellipse-parameterization (§§ computing centroid, covariance G_cov, eigenvalues G_major/G_minor, orientation G_θ) coincides with the narrower reading, so a defendant should not assume the broader construction will survive in every forum.

---

2. Level of ordinary skill in the art (LOIS)

A person having ordinary skill as of January 1998 would hold a B.S./M.S. in electrical engineering or computer science with 2–4 years' experience in capacitive touch sensing and/or real-time image/pattern recognition, and would be familiar with (a) capacitive touchpad drive/sense circuitry, (b) row/column electrode scanning and centroid interpolation, and (c) the standard image-processing pipeline: smoothing → thresholding → local maxima → connected-component segmentation → moment/covariance feature extraction → nearest-neighbor tracking. The Handbook of pattern recognition as of 1998 (Ballard & Brown; Horn; the Hu moment-invariant literature) is squarely within this artisan's ordinary toolkit. That overlap is the engine of the obviousness case.

---

3. Scope and content of the prior art

3.1 The primary reference applied by the Examiner

Bisset is the closest art, and it was of record and relied upon. It supplies: proximity sensing of multiple fingers, generation of a pixilated "finger profile," and downstream transmission of cursor/control signals.

3.2 Secondary references of record relied upon by the Examiner

3.3 Capacitive-array / scanning art cited in the specification (and in the References Cited list)

3.4 Non-patent literature of record ("Other Publications")

• S. Lee, "A Fast Multiple-Touch-Sensitive Input Device," University of Toronto Masters Thesis (1984) — a 32-row × 64-column multi-touch electrode array; capacitive changes read singly or in rectangular groups. This is the seminal multi-touch imaging reference and is expressly discussed in the '828 specification.
• Chang, C.-C. et al., "A Hashing-Oriented Nearest Neighbor Searching Scheme," Pattern Recognition Letters 14(8):625–630 (1993) — nearest-neighbor search, i.e., the matching primitive underlying contact tracking.
• Crowley, J.L., "Vision for Man-Machine Interaction," Robotics and Autonomous Systems 19(3–4):347–358 (1997).
• Davis, J. et al., "Recognizing Hand Gestures," ECCV (1994).
• Heap, T. et al., "Towards 3D Hand Tracking Using a Deformable Model," Int'l Conf. Automatic Face and Gesture Recognition (1996).
• Nirei, K. et al., "Human Hand Tracking from Binocular Image Sequences," IEEE IECON (1996).
• The specification also cites S. Alimad, "A Usable Real-Time 3D Hand Tracker," Asilomar Conf. (1994) and Y. Cui and J.… (truncated in the OCR) for free-space hand tracking.

These NPL items establish that segmenting a hand image into parts, parameterizing each part by shape/orientation, and tracking parts across successive frames was routine computer-vision practice well before 1998 — and that the artisans of the era read the vision literature when building hand-input devices. This is the single most important fact for the motivation-to-combine analysis.

3.5 What the art does not of record disclose

I found no reference of record that expressly teaches computing eigenvalues and eigenvectors of a covariance matrix associated with a segmented pixel group. That is precisely the gap the Examiner identified (see § 4.2). It is the only limitation in the '828 patent that was ever held allowable over the cited art. Any § 103 challenge must either (a) supply that teaching from art outside the record, or (b) prove it was so standard as to be within the ordinary skill level.

---

4. The prosecution record: what the Examiner actually found

This is the most probative Graham evidence available, because it is the Office's own contemporaneous § 103 analysis.

4.1 The rejections in the '828 file (application 11/677,958)

4.2 The Examiner's stated motivations to combine (verbatim reasoning, related '828-family application 11/428,521, same art)

The Examiner articulated the Bisset + Shieh rationale as follows:

"It would have been obvious to one having ordinary skill in the art at the time the invention was made to have been motivated to combine the hand recognition and multiple input device as taught by Shieh into the multiple input device of Bisset because both are within the same field of endeavor and also because of the commonly understood need to improve and increase productivity of the touch device in order to meet or exceed that of the input devices available (see Bisset at least at column 2, lines 8–14)."

And on finger/thumb differentiation (claim 4):

"It would have been obvious to one having ordinary skill in the art… to have included the concept of differentiating a finger from a thumb of Shieh in the device of Bisset in order to trigger different functions based on which finger has been used…"

These are textbook KSR rationales: same field of endeavor, known need, predictable result.

4.3 The amendment that created the wedge

Applicant responded by adding "mathematically" to claims 1 and 10:

"Nonetheless, claim 1 has been amended to recite mathematically fitting an ellipse to at least one of the pixel groups. During the interview, the Examiner indicated that the amendment would overcome the rejections."

So: the prior art did reach everything except (i) the word "mathematically" as a limitation, and (ii) the covariance eigen-decomposition of claims 5/9/15/20. Bisset disclosed everything else, including the "finger profile" that the Examiner read on the pixel-group/ellipse-parameter limitations. That is a very narrow escape.

---

5. Grounds of obviousness

Each ground below is stated with (a) the combination, (b) the claim limitations it reaches, and (c) the articulated motivation.

---

Ground 1 — Bisset US 5,825,352 alone

Reaches: the device/method structure of claims 1 (background), 10, 16, 24, 31; and directly, per the Examiner, claims 10–13, 16–19, 21, 22, 30, 33–35.

Mapping. Bisset's finger profile is a scan of a plurality of electrodes producing a set of capacitance values ("pixels") representing proximity of a touch object → producing a proximity image. Bisset identifies multiple simultaneous finger contacts in the profile → segmenting into groups, each representing a touch object. Bisset's microcontroller forms "X-Y cursor data, and control signals" and transmits them to a PC over PS/2 / RS-232 / ADB → transmitting the change in one or more ellipse parameters as a control signal, host communication interface, and the "integral with the electronic device" limitations (Examiner took Official Notice).

Motivation: single-reference obviousness — no combination needed. Under KSR, "the improvement of one reference is obvious when the improvement is a predictable use of prior-art elements according to their established functions."

Caveat: After the "mathematically" amendment, Bisset alone is no longer sufficient to invalidate claim 1 or claim 10 as issued, because the Examiner conceded that treating raw capacitance measurements as an "ellipse fit" was the one reading the prosecution history forecloses. Ground 1 therefore survives as an obviousness ground for the dependent claims that do not carry the "mathematically fitting" limitation (e.g., the transmitting/control-signal and device-integration claims) — but it cannot, standing alone, carry claim 1.

---

Ground 2 — Bisset + a moment-based (covariance) shape-parameterization reference

Reaches: claims 1, 2, 3, 23–29 (the "mathematically fitting an ellipse" limitations).

The gap: Bisset produces a finger profile but never fits an ellipse to it. What is missing is a step that converts a segmented blob into a small set of numerical shape parameters.

The secondary teaching: Central-second-moment ("moment of inertia") ellipse fitting — compute the blob's centroid, form the 2×2 covariance matrix of pixel coordinates weighted by intensity, and take its eigenvalues/eigenvectors to obtain major axis, minor axis, and orientation — is a closed-form, textbook technique of 1980s–90s image analysis. It appears in the general image-processing literature (Ballard & Brown, Computer Vision, 1982; Horn, Robot Vision, 1986; Hu's moment invariants, 1962) and in the conic-fitting literature (Bookstein, "Fitting conic sections to scattered data," 1979). Critically, the '828 patent itself presents this as the standard method — equations 15–21 compute the covariance matrix G_cov, its eigenvalues λ0, λ1, and the orientation G_θ = arctan((λ0 − G_xx)/G_xy). An inventor's own specification describing a technique in closed-form algebraic terms, without asserting it as novel, is powerful evidence of what the artisan regarded as conventional.

Motivation to combine (explicit):

1. Same field of endeavor / analogous art. Both Bisset's touchpad signal processing and second-moment shape analysis solve the same problem: reducing a set of measured intensities into compact position/shape descriptors. The Chang nearest-neighbor paper in the '828 record confirms that the touch-input community was importing pattern-recognition algorithms of exactly this kind.
2. Known problem, known solution. Once the artisan had multiple blobs (Bisset) that could be fingertips, thumbs, or palms, the artisan needed size/orientation/eccentricity to tell them apart — the specification itself says as much ("The total group proximity, eccentricity, and orientation are used by higher level modules to help distinguish finger, palm, and thumb contacts"). Second-moment ellipse fitting is the obvious, closed-form answer; it requires only a centroid and a 2×2 covariance, both trivial at the scan rates involved.
3. Predictable result. Computing an ellipse from a blob's moments yields, deterministically, the blob's approximate shape and orientation. There is no unpredictability to rebut.
4. Design incentive / market pressure. Bisset itself states the need to match or exceed the productivity of existing input devices (Col. 2, ll. 8–14) — i.e., a concrete commercial driver to add discriminating shape parameters.

Result: claims 1, 2, 3, 23–29 would have been obvious.

---

Ground 3 — Bisset + Shieh US 5,748,184

Reaches: claim 4 (and its device counterpart), i.e., using ellipse parameters to distinguish a fingertip from a thumb.

This is the Examiner's own ground, and it is the strongest one in the file because it is unrebutted on the merits: Shieh expressly distinguishes a thumb (216 → FOCUS) from an index finger (257 → OPEN), and FIG. 5 depicts the hand contacts in a ring-like formation. The Examiner's rationale — "to trigger different functions based on which finger has been used" — is a KSR "known need / predictable result" rationale, and Shieh and Bisset are both touch-input references (same field of endeavor).

---

Ground 4 — Bisset (or Ground 2/3 combination) + the covariance eigen-decomposition teaching — the crux

Reaches: claims 5, 9, 15, 20 (computing one or more eigenvalues and one or more eigenvectors of a covariance matrix associated with the pixel group).

This is the only limitation the Examiner ever found allowable. It was allowed not because the Examiner found it inventive in principle, but because no reference of record taught or suggested it. That is a § 102/§ 103 evidentiary gap, not necessarily a merits gap.

Argument:

1. Eigen-decomposition of a covariance/scatter matrix is the standard method for extracting the principal axes of a 2-D intensity distribution. The '828 specification does not claim to have invented it; it recites the algebra (equations 15–21) as the natural implementation.
2. Ground 2 already supplies moment-based ellipse fitting. Computing eigenvalues and eigenvectors is the moment-based ellipse fit — the eigenvalues of the covariance matrix are the squared semi-axis lengths and the eigenvectors are the axes (the specification says exactly this: G_major = √λ0, G_minor = √λ1, G_θ = arctan(...)). So the additional limitation of claims 5/9/15/20 adds nothing beyond the mathematics that makes the Ground-2 ellipse computable from moments.
3. Motivation: numerical stability, closed-form computation, and the need for orientation and eccentricity (not merely position) to discriminate hand parts. A 2×2 symmetric covariance matrix has a trivial closed-form eigen-decomposition; the artisan would use it because it is the obvious, cheapest, and most numerically robust route.

Risk flag — this is the ground most likely to fail. The Examiner's "allowable subject matter" statement is an express finding, on the face of the record, that the prior art does not teach the eigen-decomposition. An invalidity challenger must therefore present a specific reference (a moment-analysis text or arm of the art) that teaches covariance eigen-decomposition for blob/shape characterization; merely arguing "it's routine" invites the In re Kahn / "conclusory reasoning" objection. Also note KSR does not relieve a challenger of the duty to identify where each limitation is taught.

---

Ground 5 — Adding the tracking layer

Reaches: the claims reciting time-sequenced images and tracking a path/change through successive images (claim 2 family; claims 31, 32 and the means-plus-function equivalents).

Combination: Bisset (or Ground 2) + Chang, "A Hashing-Oriented Nearest Neighbor Searching Scheme" (1993) (of record) and/or the of-record hand-tracking NPL (Crowley 1997; Heap 1996; Nirei 1996; Davis 1994) + Lee's 1984 FMTSID thesis.

Mapping & motivation: Lee discloses a scanned multi-touch electrode array producing successive proximity frames; Chang discloses nearest-neighbor matching — the exact operation the '828 specification uses to match current groups to predicted path positions ("a group and path… will be closer to one another than to other groups and paths"); the hand-tracking NPL (Heap's deformable models, Nirei's binocular tracking) discloses tracking hand parts across successive frames. The motivation is direct and KSR-apt: once you can image multiple contacts, tracking them is the necessary next step to derive velocity/control signals — and Bisset already teaches deriving motion/control signals.

---

Ground 6 — The means-plus-function claims (24–35)

Claims 24–35 invoke "means for…" language. Under § 112(f), these are limited to the corresponding structure disclosed in the specification (plus equivalents) — scanners, segmentation code, ellipse-fitting code, and the tracking/communication modules. That narrows the claims, which generally helps an invalidity challenger: the challenger need only show the corresponding structure was known. Conversely, a § 112(f) indefiniteness attack (Aristocrat Technologies v. International Game Technology, 521 F.3d 1328 (Fed. Cir. 2008) — computer-implemented means-plus-function claims require disclosure of a specific algorithm) is available: if the segmenting/fitting "means" are claimed purely functionally without a disclosed algorithm, the claims may be indefinite. The file history shows Applicant actually raised the Examiner's failure to make a § 112 ¶ 6 finding as a procedural objection — meaning the § 112(f) issue is squarely present in the record.

---

6. Consolidated motivation-to-combine summary

---

7. Counterarguments — objective indicia and record weaknesses

A responsible § 103 opinion must present the other side. The '828 patent has significant objective evidence, and it was actually litigated:

1. Validity was upheld at the ITC. The ALJ found the asserted '828 claims (1, 2, 10, 11, 24–26, 29) not invalid, and the Commission's notice of review (77 Fed. Reg. 16,842, Mar. 22, 2012) reflects that the validity finding was not disturbed. The Federal Circuit's 2013 opinion vacated and remanded on infringement/claim construction; it did not reverse the validity holding. A challenger today must overcome a prior adjudication of validity — a heavy practical burden even though not issue-preclusive against a non-party.
2. Real technical differences. The Examiner's own allowance of claims 5/9/15/20 is a documented finding of a genuine gap over the art. Any § 103 case must close that gap with new art.
3. Commercial success, copying, industry praise. Apple argued (and the Federal Circuit credited in the parallel '607 analysis, remanding specifically so the ITC could weigh secondary considerations) that the multitouch touchscreen "was largely responsible for the praise, copying, and commercial success of the iPhone." Google's own conduct — re-enabling pinch-to-zoom in Android 2.1 after Apple's assertions — is a copying/nexus data point. These are the Graham factor-(iv) facts that sunk the ITC's obviousness analysis on the '607 patent in 2013.
4. Long-felt need / failure of others. The specification's framing — that no prior device integrated typing, pointing, scrolling, and handwriting, and that resting hands confused prior touch devices — supports a long-felt-need narrative.
5. § 325(d) / Becton Dickinson exposure. Bisset, Shieh, and Tareev are all of record and were relied upon. Any forum (PTAB or district court) will scrutinize a re-run of the same art; the Becton, Dickinson factors weigh against a challenge that merely recycles the Examiner's references with a new secondary reference.
6. Expired status moots the practical stakes. The patent expired (adjusted) on 2021-01-01, status "Expired – Fee Related." There is no live royalty base and no injunction exposure, which means the usual litigation incentives to fund a full invalidity case are absent.

---

8. Bottom line

• Strongest obviousness grounds (high confidence, on the face of the record):
  • Bisset US 5,825,352 alone — for the claims without the "mathematically fitting" limitation (the transmitting/control-signal and device-integration claims the Examiner rejected under § 103(a): 10–13, 16–19, 21, 22, 30, 33–35).
  • Bisset + Shieh US 5,748,184 — for the fingertip/thumb discrimination claim (Examiner's own ground; unrebutted).
• Defensible but art-dependent grounds (medium confidence):
  • Bisset + moment-based ellipse fitting (general image-processing texts of the 1980s–90s) → claim 1/10/23–29.
  • + Chang (1993) and the of-record hand-tracking NPL → the tracking claims.
• Weakest ground (the one to attack or to shore up):
  • Covariance eigenvalue/eigenvector claims (5, 9, 15, 20). This is the only limitation ever held allowable, and the Examiner expressly found the prior art silent. A challenger needs a specific teaching of covariance eigen-decomposition for shape characterization; a bare "it was routine" argument will not carry.
• Claim construction matrix risk: the Federal Circuit rejected the ITC's narrow "actually fitted" construction in Apple's favor. The broader the construction, the easier the obviousness case — but the narrower reading (which the specification's own equations 15–21 support) is what a challenger should assume for planning purposes, because it is the reading most favorable to the patentee.
• No new trial milestones: as established in the PTAB section, there is no AIA proceeding, therefore no § 315(e)(2) estoppel, and no FWD to borrow. Any § 103 case must be built on a fresh record.

Confidence statement. The prosecution-history facts (the specific rejections, the Bisset/Shieh/Tareev citations, and the "allowable subject matter" finding on the covariance limitation) are high confidence — they are quoted from file-wrapper exhibits. The status of the ITC/Federal Circuit construction is high confidence that the earlier section's narrow-construction premise is wrong, based on three independent secondary sources; I did not retrieve the Federal Circuit opinion text verbatim in this session, so the precise scope of the Fed Cir's substituted construction is medium confidence. The attribution of specific secondary references to the moment/eigen-decomposition teachings (Ballard & Brown; Horn; Bookstein; Hu) is training-data based and NOT verified against the '828 record in this session — those references are not in the '828 References Cited list, and I flag them as candidate art to be verified before being relied upon in any pleading.