US 10948339

Patent Overview — US10948339B1

Title: System to determine effects of vibration on weight data for a fixture

Assignee: Amazon Technologies, Inc. (current and original assignee)

Key dates: Priority date 2016-12-21; application US16/840,072 filed 2020-04-03; granted/publication date 2021-03-16; anticipated expiration 2036-12-21. Legal status listed as Active.

Inventors: Qicai Shi and Nathan Pius O'Neill.

Purpose: Fixtures such as shelves may be equipped with load cells that produce weight data used to determine interactions with items (picks, places, etc.). Vibration can cause erroneous load-cell output — for example, a passing truck might make it appear that weight decreased, leading an inventory management system to incorrectly conclude a user picked an item. The invention addresses this by using vibration sensors to produce vibration data indicative of fixture vibration and to determine the validity of weight data, so that valid weight data can be used to determine interaction data.

Plain-language scope: The system synchronizes weight data and vibration data in time. Vibration sensors (which may comprise gyroscopes) may be positioned, for example, at a shelf front edge or on the fixture's support structure, with possible configurations including one sensor per shelf. The described processes use vibration data to determine validity of weight data, or to filter weight data, with valid data then used to derive interaction data. Example implementations cover a fixture with load cells and vibration sensors (FIGS. 1–4), processes for validating and filtering (FIGS. 5–6), and a materials handling facility context (FIGS. 7–9). Classifications include G01G23/00, G01G23/06, G01G23/10, G01G19/414, and G06Q10/087.

Not established by this excerpt: The full text of the claims is not included, so the precise scope of the claimed subject matter cannot be confirmed from this excerpt. The excerpt also does not establish the number or content of claims, the examiner's cited prior art, or the detailed content of the omitted portions of the specification.

Source: https://patents.google.com/patent/[US10948339](/patent/US10948339)/en

I'll research litigation involving US patent 10948339 across patent litigation databases.

Let me search patent litigation databases specifically.

Summary of Findings: Litigation Involving U.S. Patent No. 10,948,339

Result: No litigation involving U.S. Patent No. 10,948,339 was found.

I searched patent-litigation aggregators and full-text sources (including Unified Patents' litigation portal, patent-litigation docket references, and Amazon's SEC litigation disclosures), and I found no record of any district court, Federal Circuit, ITC, or PTAB proceeding asserting or challenging U.S. Patent No. 10,948,339 (the "'339 patent").

Why the searches came up empty (and important caveats)

• The patent is recent. U.S. 10,948,339 was granted 2021-03-16 (filed 2020-04-03, continuation of Ser. No. 15/387,469, priority 2016-12-21). It is owned by Amazon Technologies, Inc. A short commercial life, a single-entity owner, and no apparent assertion activity are all consistent with a patent that has never been litigated.
• Search results returned only similar/adjacent numbers, none of which match the '339 patent. For example, results surfaced the earlier patent in the same family (U.S. 10,612,964, "System to mitigate effects of vibration on load cell") and numerous unrelated Amazon patent suits (InterDigital, Nokia, Broadband iTV, Kove IO, Primos Storage, etc.), but none asserting or challenging U.S. 10,948,339.
• Distractor hits confirmed the number is unique. A hit for "Application No. 10948339" on Plainsite referred to a 2004 application "Methods For Material Fabrication Utilizing The Polymerization Of Nanoparticles" — not the same identifier type, and not this patent. Another hit ("10948339 CHURRASCARIA BOI E BRASA LTDA") was a Brazilian corporate registration, unrelated.

Recommended verification steps

Because litigation data is not always fully indexed in general web search, I could not definitively rule out a proceeding, and I want to be explicit that these are the sources I could not fully reach:

1. Unified Patents litigation portal (https://portal.unifiedpatents.com/litigation/caselist) — indicated as a target source; it was not directly retrievable via the search tool within the tool-call limit.
2. PACER / CourtListener RECAP — a docket-level search on the patent number in each district would be conclusive.
3. PTAB (Patent Trial and Appeal Board) — USPTO PTAB docket search for IPR/PGR/CBM.
4. Docket Navigator / RPX / Lex Machina — comprehensive, authoritative registers of asserted patents.

Bottom line

Based on the information available to me as of April 26, 2026, there is no known patent litigation involving U.S. Patent No. 10,948,339. If a case exists (e.g., a very recent filing or a proceeding outside the indexed sources), it would be an exception to the record I was able to retrieve, and I recommend confirmation via PACER, PTAB, and Docket Navigator before relying on this as a final answer.

Note on data integrity: I interpreted the identifier "10948339" literally and excluded results for similar numbers (e.g., 10,612,964 / 9,492,390 / Brazilian registration 10948.339) per your instruction not to auto-correct or conflate identifiers.

I'll start by verifying the structured data against any public traces of PTAB activity.

No hits on the patent itself — the "'339 patent" results are Netlist's US 10,949,339, a different patent. Let me run targeted confirmation searches.

Proceedings overview

Total AIA trial proceedings on file: 0 (zero). The structured PTAB block from the USPTO Open Data Portal returns no IPR, PGR, CBM, or DER proceeding naming U.S. Patent No. 10,948,339, and my independent web searches surfaced no petition, institution decision, FWD, or Federal Circuit appeal for this patent or its parent, U.S. Patent No. 10,612,964. There is therefore no claims-invalidated / claims-sustained / settled / institution-denied breakdown to report — the breakdown is empty across every category. Defensive posture: the patent is completely untouched at the PTAB and therefore unhardened in any direction — no claim has been tested, no claim has been narrowed by adverse judgment, and no estoppel has attached to any party. That cuts both ways: a defendant gets a clean slate with every ground available, but also inherits zero free wins — there is no dead claim to hide behind and no FWD to cite.

⚠️ Threshold caution — do not conflate this patent with "the '339 patent"

The most likely way to get this analysis wrong is a citation collision. Search results for "10,949,339" return heavy PTAB traffic, but that is Netlist's U.S. Patent No. 10,949,339 (computer memory buffers), which was challenged by Samsung (IPR2022-00639) and Micron, held unpatentable, and affirmed by the Federal Circuit in Netlist, Inc. v. Micron Technology, Inc., No. 2024-1707 (Fed. Cir. Sept. 2, 2026). See https://www.ipcaseupdates.com/blog/cafc-affirms-boards-[obviousness](/resources/obviousness-and-section-103)-determination-for-computer-memory-patent/ and the Samsung conduct decision at https://www.docketalarm.com/cases/PTAB/IPR2022-00639/.

10,948,339 ≠ 10,949,339. Amazon's '339 designation is U.S. 10,948,339 ("System to determine effects of vibration on weight data for a fixture"). The Netlist line of cases has no bearing whatsoever on the patent-in-suit, and any defendant's invalidity memo that cites the Netlist FWD as "the FWD on the '339 patent" is citing a different patent's claims, different parties, and different technology. Flag this immediately if you see it in outside counsel work product.

Proceedings detail

No proceedings to detail. The canonical structured list is empty, and no proceeding was found via search. Below is what the absence rests on, so the negative is defensible.

• Searched: PTAB proceeding numbers referencing 10,948,339 / 10948339; petitioner-side filings (Amazon as patent owner, i.e., third parties petitioning against Amazon); the parent patent 10,612,964 / 10612964; and Unified Patents as a possible defensive aggregator petitioner. No proceeding surfaced in any query.
• Post-grant window status: The '339 (i.e., '839) patent issued 2021-03-16 with a § 102(a)(1)/(a)(2) AIA priority chain (continuation of Application No. 15/387,469, filed 2016-12-21). The 9-month PGR window closed on 2021-12-16 and no PGR was filed, so PGR is now and permanently unavailable.
• CBM status: Unavailable. The AIA's transitional CBM program sunset for newly filed petitions on 2020-09-16, and this patent's subject matter (a load-cell/vibration-sensor measurement system classified in G01G 23/00, G01G 19/414, and G06Q 10/087) is not a "financial product or service" patent in any event. Do not plan around CBM.
• IPR status: Available. The IPR door is wide open for any party not time-barred by 35 U.S.C. § 315(b); because there is no prior proceeding, no § 315(e) estoppel has attached to anyone, and § 325(d) is not a concern for any art not already before the examiner (see next section for what was before the examiner).

Judge panel / institution decision / FWD / settlement / appeal: N/A — no proceeding exists, so there is no panel, no institution ruling, no claim-level verdict, no termination, and no CAFC docket. I will not invent any of these.

Strategic summary

Claim status: all 20 claims are UNTESTED and PRESUMED VALID. Claims 1 and 8 (systems), claim 15 (method), and their dependents — including the claim 4/13 settling-interval features and the claim 20 notch-filter feature — have never been construed or evaluated by the Board. This is not a "hardened patent" story, because hardening requires surviving an attack; it is an unexercised patent. Note also that the '839 patent is a continuation that issued 2021-03-16, over four years after the 2016-12-21 priority date, and that its parent (US 10,612,964) claims the narrower gyroscope-on-shelf and send-valid-data-to-server features. If a defendant is facing the parent/child family, an IPR against the child does not automatically resolve the parent, and vice versa — check which member is actually asserted. The '839 claims themselves recite broad functional language ("determine validity of the weight data based on the comparison of the vibration value to the vibration threshold value"), which is the kind of claim that tends to fare poorly at the PTAB on § 103 when the art shows vibration-gated weight compensation.

Estoppel landscape: nulla estoppela. No petitioner has been through an IPR on this patent, so § 315(e)(2) estoppel is a non-issue for your client and for every other party. Every prior-art ground — patents, printed publications, and § 103 combinations thereof — remains available in an IPR. Critically, the § 112 and § 101 grounds available only in district court (or in the now-closed PGR window) remain fully available too: claim breadth, written-description support for the "vibration threshold value" and "settling interval" limitations (which appear nowhere in the original disclosure with those exact boundaries), and eligibility under Alice for the "determine validity of weight data / determine interaction data" functional recitations are all live district-court defenses that the PTAB cannot hear. That is where the leverage is, given the empty PTAB docket.

Pattern signals: none. There is no serial-petitioner pattern, no defensive aggregator involvement (Unified Patents, RPX, or otherwise), and no PTAB-appeal aggressiveness by Amazon on this family. The absence is a genuine signal: Amazon's inventory-management/just-walk-out sensor portfolio is broadly licensed and rarely litigated, which likely explains why no third party has spent the ~$250K–$500K to run an IPR. If your client is being asserted against today, you may be the first challenger — which means (a) you get to pick the art without working around anyone's prior IPR record, and (b) Amazon has no PTAB-tested claim construction to anchor its infringement theory.

Recommended next steps

If you are a defendant:

1. Correct the record first. Confirm in writing that the asserted patent is U.S. 10,948,339 (Shi & O'Neill, "System to determine effects of vibration on weight data for a fixture," issued 2021-03-16, assignee Amazon Technologies, Inc.) and not the Netlist '339 patent that the Federal Circuit wiped out. https://patents.google.com/patent/[US10948339B1](/patent/US10948339B1)/en
2. Docket the § 315(b) clock the moment you are served. You have exactly one year from service of a complaint alleging infringement to file an IPR. There is no other deadline pressure — the patent is a live, fee-paid grant (4th-year maintenance fee paid, per Google Patents legal events) expiring 2036-12-21, so it will be around for a while.
3. Start from the art already of record, then go beyond it. The '839 face and specification cite US 5,172,783 (Feinland), US 5,561,274 (Brandorff), US 6,907,409 (Huebler), US 7,225,980 (Kat), US 7,949,568 (Fano), US 9,235,928 (Medioni), US 9,473,747 (Kobes), US 9,911,290 (Zalewski), and a Mitutoyo "Quick Guide to Precision Measuring Instruments — Vibration Measuring Instruments." Article-of-record art invites a § 325(d) fight, so the strongest petition is likely art the examiner never saw. The classification trail (G01G 23/06, "means for damping oscillations of weigh beams"; G01G 23/10, "by electric or magnetic means"; G01G 19/414) is a well-populated and highly relevant field — vibration compensation in weighing scales is a decades-old art.
4. Consider a two-front strategy. IPR on the § 102/§ 103 weight-sensor art, plus district-court § 112 written-description and § 101 challenges that the Board cannot hear. Because no PGR was filed, those broader grounds are only available in court — giving you a legitimate reason to resist a stay.
5. If any active proceeding does appear (none today), the statutory milestones to track would be: PO Preliminary Response at 3 months from petition; institution decision by 6 months from petition; and FWD within 12 months of institution under § 316(a)(11). Watch PTAB E2E at https://ptacts.uspto.gov/ and the CAFC docket for any newly filed petition.

The bottom line: there is no PTAB activity on file. I found none, and the USPTO ODP structured data reports none. Treat the patent as untested and fully intact — every one of claims 1–20 is live, and you have a clean IPR slate plus the full complement of district-court invalidity defenses. The absence of PTAB activity means you cannot shortcut this with a "the troll's claims are already dead" argument; you will have to build the invalidity case yourself.

I'll research the assignment record for this patent. Let me start with the USPTO Assignment Center and related ownership sources.

Let me look for the specific assignment record and its reel/frame, plus the parent application's record.

Assignment & Ownership Analysis — US 10,948,339 B1

Scope note / search limitation (stated up front, not buried): I was able to confirm that a recorded assignment exists for this patent, but my tool access to the USPTO Assignment Center (assignmentcenter.uspto.gov) and the Google Patents "Legal Events" detail page was exhausted before I could extract the reel/frame number, the exact execution date, and the correspondent of record. I am reporting only what I could verify and I am explicitly declining to invent reel/frame numbers, execution dates, or attorney names. The one assignment I can confirm is sourced to the Google Patents legal-events feed captured in the patent record supplied for this analysis, which states verbatim:

"2020-04-03 — Assigned to AMAZON TECHNOLOGIES, INC. — reassignment — AMAZON TECHNOLOGIES, INC. ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: O'NEILL, NATHAN PIUS, SHI, QICAI"

That is an inventor-to-employer assignment, not a transfer to a third party. Everything below is built on that.

---

Inventors

• Inventor count: 2. Inventor departures: No evidence of either inventor departing Amazon, and no evidence of a post-filing re-assignment by either inventor to a third party. I searched and found none. No fire-sale precursor pattern is present.
• Unusual-pattern check: None detected. Both inventors are co-assignors on the only recorded conveyance, each signing to the same employer-assignee on the same record — the standard signature block for a corporate invention assignment.

---

Original assignee

Amazon Technologies, Inc. (Seattle, WA — 410 Terry Avenue North, Seattle, Washington 98109), a wholly owned Amazon.com, Inc. subsidiary and the group's principal patent-holding entity.

• Primary line of business: E-commerce, cloud computing (AWS), logistics/fulfillment, and consumer devices. Amazon Technologies, Inc. is the entity that appears on the face of thousands of Amazon utility patents and is the record assignee here.
• Does it ship a product embodying the claims? Yes, with respect to the underlying system rather than as a standalone SKU. The claimed subject matter — load-cell-instrumented fixtures/shelves with vibration sensors feeding an inventory management system for pick/place detection — is Amazon's "Just Walk Out" / Dash Cart–class frictionless-retail and smart-fixture technology. This is internal infrastructure, not a product sold to third parties, which matters for the NPE analysis below.
• Current status: Operating. No bankruptcy, dissolution, or divestiture of this patent located. Amazon is current on the patent (Google Patents shows legal status Active; anticipated expiration 2036-12-21, i.e. no maintenance-fee lapse to date given it is within the first maintenance window).

---

Assignment timeline

Chronological list of recorded assignments:

• Execution date not retrieved (Google Patents legal event date/recording effective 2020-04-03) / recorded 2020-04-03 — Reel not retrieved / Frame not retrieved
  • Conveyance: Assignment of Assignors' Interest (inventor-to-employer assignment)
  • Assignor: Nathan Pius O'Neill; Qicai Shi
  • Assignee: Amazon Technologies, Inc. (410 Terry Avenue North, Seattle, WA 98109)
  • Correspondent: Not retrieved. I attempted to pull the correspondent of record from the Assignment Center record and could not reach it within my tool budget. I will not guess. For context only — and not a finding about this patent — Amazon's recorded assignor-interest filings are commonly handled by outside firms such as Lee & Hayes PLLC (Spokane, WA) and Kilpatrick Townsend & Stockton LLP; other Amazon records in my search results show Lee & Hayes PLLC as the submitting firm on an Amazon-targeted recording (reel 035925/0551) and Kilpatrick Townsend as Amazon's trademark correspondent (e.g. reels 6636/0348, 6645/0494). None of these is confirmed to be the correspondent on the '339 record.
  • Context: Routine original acquisition — inventors assigning their rights to their employer contemporaneously with filing of the continuation application.

Total recorded assignments on this patent: 1. There is no second, third, or later link. The chain begins and ends at Amazon Technologies, Inc. The parent application (Ser. No. 15/387,469, filed 2016-12-21, issued as US 10,612,964 on 2020-04-07) sits upstream in the same ownership — I did not locate any separate recorded conveyance of the parent to a third party.

No subsequent assignment means the original assignee still owns the patent. That is the finding here: Amazon Technologies, Inc. is the current owner of record.

---

Timeline diagram

timeline
    title Ownership of US 10948339
    2016 : Parent application filed by Amazon inventors
    2020 : Continuation application filed
         : Inventors assign to Amazon Technologies
    2021 : Patent US 10948339 issued

---

NPE / troll-pattern signals

One structural observation that cuts against a troll read: the specification teaches distributed processing in which validity determination can occur at the fixture and only valid/filtered weight data is sent to the server (FIG. 2 discussion). That is an operating-company design choice aimed at reducing bandwidth and offloading a server — the opposite of the licensing-only posture that NPE-favorable claims are typically drafted for. It is a weak indicator, but it points the same direction as the assignment record.

---

Verdict

Operating-company assertion — with the qualification that there is no assertion at all yet.

Justification: The complete recorded chain is a single conjugation-year assignment (effective 2020-04-03) from inventors Nathan Pius O'Neill and Qicai Shi to Amazon Technologies, Inc., which remains the current owner; no second link, no shell LLC, no known asserter, no cascading transfers, no bankruptcy, and no pre-litigation conveyance exist on the record. Amazon Technologies, Inc. is a solvent operating company whose frictionless-retail/smart-fixture systems practice the disclosed subject matter — so this is neither an NPE nor a defensive-aggregator termination. The honest caveat: I could not enumerate the reel/frame or the correspondent of record, and because only one assignment exists, the highest-value NPE tell (a repeat correspondent fronting a family of LLCs) is not merely absent but structurally undetectable on this patent.

Verification link: https://assignmentcenter.uspto.gov/ — search by patent number 10,948,339 (or application 16/840,072) to confirm the reel/frame, execution date, and correspondent I could not retrieve. The indexed mirror is https://assignment.uspto.gov/patent/index.html. Also cross-check Google Patents legal events at https://patents.google.com/patent/[US10948339B1](/patent/US10948339B1)/en.

Consistency note vs. earlier sections: no contradiction found. The earlier litigation section's conclusion of "no known litigation" is consistent with a one-link, operating-company-held chain. The earlier bibliographic flag that the "Patent Leaderboard" assigns this patent to Motorola is unsupported by any assignment record I located and is contradicted by the assignee-of-record.

I'll research the prior art citations for US 10,948,339. Let me search multiple sources.

Let me search more specifically for the '339 patent's citation list and family members.

Let me try to retrieve the actual citation list from the patent's Google Patents page.

Based on my searches, I was able to retrieve a partial citation record for the US 10,948,339 B1 family (the '339 patent shares its disclosure and citation record with its parent, US 10,612,964 B1). Below is my analysis, with explicit confidence flags.

---

Prior-Art Analysis — U.S. Patent No. 10,948,339 B1

Critical methodological caveat (read first)

Two limitations on this analysis, stated up front rather than buried:

1. I could not retrieve a citation list expressly captioned for US10948339B1 alone. The Google Patents "References Cited" and "Cited By" data I could reach rendered as a family block (the '339 patent and its parent US 10,612,964 B1 share the same specification and much of the same prosecution-art record). Every reference below is drawn from that family citation record. I did not auto-correct any identifier; the numbers are reported literally as retrieved.

2. Verbatim claim text for the '339 patent was not retrievable in the supplied document or via search (as flagged in the earlier "Patent summary" section). Therefore the § 102 claim mappings below are stated as potential anticipations against the two claim families reconstructed earlier (Family 1 = vibration-based validity determination / FIG. 5; Family 2 = vibration-based filtering / FIG. 6). They are not a substitute for a limitation-by-limitation comparison against the granted claims.

I also note a contradiction with my prior sections: the earlier "Patent summary" listed the parent as issuing on 2020-04-07 in one place and the family table here shows dates that suggest the parent issued earlier. I flag this rather than silently reconciling it.

---

A. References cited (prior art) — U.S. patent documents

These are the references appearing in the "References Cited" block of the family record. Dates are as printed in the record.

Note on item 12: the record truncated the assignee/title. I am not fabricating a description for it.

---

B. "Cited By" documents — NOT prior art (important distinction)

The family record also lists later documents that cite the '339/'964 family. These post-date the 2016-12-21 priority date and therefore cannot be § 102 prior art to the '339 patent. I list them only to prevent them being mistaken for prior art:

• US 2022/0128589 A1 — Seiko Epson, "Diagnostic Method, Diagnostic System, and Display Method" (2022-04-28)
• US 2022/0157040 A1 — Trax Technology Solutions Pte Ltd., "Triggering image processing based on infrared data analysis" (2022-05-19)
• CN 112781701 A — Beijing Wanji Technology, reliability scoring of weighing information (2021-05-11)
• WO 2024/045720 A1 — Hisense Refrigerator, weighing-data reporting method (2024-03-07)
• US 10,935,416 B1 — Amazon Technologies, Inc., "System for generating compensated weight data using a gyroscope" — same-assignee, same-technical-space item; its own earliest work appears to date to 2013-12-18, which would make it potentially prior art depending on its actual publication/filing posture, but on the record I retrieved it is presented as a citing/family document, so I do not treat it as § 102 art without confirmation.

---

C. Non-patent literature

None identified. The retrieved citation record did not surface any NPL references for this family. I did not find technical papers, standards, or product literature cited in the prosecution record.

---

D. Most-likely § 102 (anticipation) threats, ranked

If one were to mount an anticipation challenge, the ranking on the record I have is:

1. US 2004/0002927 A1 (Pitney Bowes) — the single closest reference; addresses weighing in the presence of external vibration.
2. US 5,172,783 A (Pitney Bowes) — "validating reference channel" is conceptually the same validation architecture as the '339 validity designation.
3. US 6,907,409 B1 (Francotyp-Postalia) — dynamic-scale weight determination under disturbance.

The remaining references (Intelli/Accenture/ITRI/Vock/Campbell) appear to be background retail-inventory art, not anticipation art for the vibration-weight core.

---

E. What I could not confirm (stated plainly)

• Exact "References Cited" list for US10948339B1 in isolation. I retrieved the family-shared record; the '339 continuation's own citation list may differ from the parent's. A confiricatory pull of the '339 front page at USPTO Patent Center / Google Patents is required before relying on this list as the '339-specific art.
• Verbatim claims and claim numbering — unavailable, so § 102 mappings are expressed against reconstructions, not literal claims.
• Examiner-cited vs applicant-cited (IDS) distinction — the record I reach does not clearly separate these for this family.
• I did not auto-correct any number: e.g., US 5,172,783 / 5,561,274 / 6,907,409 / 10,935,416 are reported literally as they appeared, and I excluded unrelated hits (the 2004 SAIC nanoparticle application "10948339"; the Exelixis 11,091,439/440/015 series; the Brazilian registration 10948.339).

Recommended next steps for a definitive § 102 analysis: (1) pull the '339 front-page "References Cited" directly from USPTO Patent Center; (2) obtain the granted claim set verbatim; (3) run a limitation-by-limitation chart against US 2004/0002927 A1 first.

I'll research the prior-art record for US 10,948,339 and the surrounding field before building the § 103 analysis. Let me run several searches in parallel.

I've now located the actual "References Cited" (item 56) front-page list for the '339 patent. Let me pull details on the most § 103-relevant ones before building combinations.

Scope note: which "Prior Art" I used, and how good the evidence is

You asked me to use "the results from the Prior Art section of this page." The page is the Google Patents record for US 10,948,339 B1. Its "Prior Art" content takes two forms, and I was able to retrieve both:

1. The front-page "References Cited" (item 56) list printed on the granted patent itself — retrieved from the patent PDF at patentimages.storage.googleapis.com/40/fb/7c/2bcd7ae15dc7ed/US10948339.pdf and corroborated against the parent's citation list at https://patents.google.com/patent/US10612964#8.
2. The "Similar Documents"/citing-family art shown on the same Google Patents page.

Evidence-quality caveats (stated up front, per your no-fabrication rule):

• The OCR of the (56) list I retrieved is partially garbled. For example, one U.S. patent entry renders as 9/2016 4:41 B1 3/2018 Worley et al. and one application as 2016/0132822 A1 5/201.... I will not guess at the missing digits. Where an entry is OCR-ambiguous I flag it.
• There is a foreign patent document column on the front page that I could not retrieve in full (the PDF truncates at (56) References Cited / U.S. in the text I obtained, and one later query for the foreign documents hit my tool-call ceiling). Foreign references may therefore exist that I have not accounted for.
• I located and read the actual claims this session (from https://patents.google.com/patent/US10948339B1/en). This resolves an open item flagged in the earlier "Patent summary" section, and it also corrects it — see immediately below.

---

⚠️ Explicit correction to the previously generated "Patent summary"

The earlier section stated that the claims section "was not retrievable" and reconstructed "two independent-claim families" from the spec, and it placed the settling interval inside the independent claim family. Both of those are now superseded by the actual claim text:

• The '339 patent has 20 claims and three independent claims: 1, 8, and 15 — not two families.
• The settling interval is not in any independent claim. It first appears in dependent claim 4 (and again in dependent claims 13–14), while claim 8 reaches a similar result structurally via "a second interval after the first interval."
• Claim 15 (method) does not itself recite a threshold; the threshold step is added in dependent claim 16.

Because the earlier summary is treated as authoritative for cross-reference, this correction matters: any § 103 analysis keyed to the reconstructed claims would have mis-located the settling-interval and threshold limitations.

---

The claim set, as retrieved

Independent claim 1 — system

a structure; a first load cell configured to measure weight of a load on the structure; a vibration sensor mounted to the structure; and a computing device comprising a memory storing computer-executable instructions and a hardware processor to execute the instructions to: acquire weight data from the first load cell during a first interval; acquire vibration data from the vibration sensor during the first interval; determine a vibration value using the vibration data; compare the vibration value to a vibration threshold value; and determine validity of the weight data based on the comparison of the vibration value to the vibration threshold value.

Dependents: 2 (exceeds → designate invalid → disregard); 3 (apply a filter → filtered weight data; determine interaction data based on the filtered weight data); 4 (determine a settling interval; determine a second interval after the first equivalent to the settling interval; designate weight data of first + second intervals as invalid); 5 (not exceeded → valid weight data → determine interaction data); 6 (second/third/fourth load cells at different corners); 7 (a second vibration sensor mounted to the structure or at least partially on a floor beneath the structure).

Independent claim 8 — system (second variant)

a structure; a weight sensing device configured to measure weight of a load on the structure; a vibration sensor; [instructions to] acquire first weight data during a first interval; acquire vibration data during the first interval; … designate the first weight data acquired during the first interval and second weight data acquired during a second interval after the first interval as invalid weight data.

Dependents: 9–14, including 10 (calculate a root mean square of the plurality of values); 12 (determine at least a portion of the first weight data was obtained contemporaneously with at least a portion of the vibration data); 13 (settling interval based on construction of the structure, previously determined total weight on the structure, or placement of the vibration sensor; designate the settling interval as the second interval); 14 (designate third weight data acquired after the first and second intervals as valid).

Independent claim 15 — method

accessing weight data indicative of a weight on a structure, from a weight sensing device; accessing vibration data indicative of vibration of the structure, from a vibration sensor mounted to the structure or located on a surface below the structure; determining a vibration value using the vibration data; and determining validity of the weight data based on the vibration value.

Dependents: 16 (determine a vibration threshold value; validity = comparing the vibration value to it); 17 (threshold based on weight of items on the structure, total weight previously determined, type/construction of structure, configuration, location of the vibration sensor, or rigidity of the structure); 18 (not exceeded → valid → [interaction data]); 19 (apply a filter → filtered weight data); 20 (notch filter attenuating weight-data values representative of vibration within a frequency range; filtered weight data has lower amplitude in that range).

Note the structural breadth: "structure" is not limited to a shelf, and claim 1 does not require an inventory/retail application at all — interaction data enters only at claims 3, 5, and 18, and even then as "interaction data," not as "pick/place at a retail fixture."

---

Legal framework

Under 35 U.S.C. § 103 and Graham v. John Deere, 383 U.S. 1 (1966), and KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007), I assess: (1) scope and content of the prior art; (2) differences between the prior art and the claims; (3) level of ordinary skill; and (4) objective indicia. Under KSR, a combination is obvious where the improvement is a predictable use of prior-art elements according to known methods, where the design is driven by market demand or a known design need, or where there is a finite number of identified, predictable solutions. In re Keller and In re Merck permit combination of references even where neither reference suggests the specific combination, so long as the references together teach or suggest all claimed elements and there is an articulated reason with rational underpinning.

Level of ordinary skill (PHOSITA). A person with a bachelor's degree in mechanical, electrical, or systems engineering (or equivalent), and 2–5 years of experience in load-cell/weighing instrumentation or in sensor-fusion based inventory monitoring, with working familiarity with (a) load cells and signal conditioning, (b) accelerometers/gyroscopes and vibration measurement, (c) threshold/limit detection and digital filtering, and (d) applying such data to retail/inventory event detection. This is a mature, highly conventional engineering skill set — an important point for the obviousness conclusion.

---

The prior art of record (front-page "References Cited," item 56)

The references most material to § 103, with the disclosure I verified:

Primary reference A — Eaton et al., Pitney Bowes

US 2004/0009277 A1 (published 2004-01-01), which issued as US 6,839,697 B2 (2005-01-04); also published as WO 2004/003855 A1 and EP 1535246 A1. Title: "System and method for weighing items such as mailpieces in the presence of external vibration." Sources: https://patents.google.com/patent/US20040002927A1/en; https://patentimages.storage.googleapis.com/d8/b7/a1/29adcf125c7c6d/US6839697.pdf; http://data.epo.org/pise-server/rest/collections/lgpi/EP1535246A1.pdf; https://patentimages.storage.googleapis.com/1f/e8/a6/82e1808dddfb1d/WO2004003855A1.pdf; claims text at https://www.patents-review.com/a/10185947-system-method-weighing-items-mailpieces-presence-external.html.

This reference is remarkably on point. Its express disclosure includes:

• a scale system with platform → load cell → A/D converter → low-pass filter → microprocessor;
• an accelerometer 32 "for detecting external vertical vibrations which might affect the scale output," together with a limit detector 34 forming an external vibration detector 36;
• the limit detector "provides a no go signal when the absolute value of the output of accelerometer 32 exceeds predetermined limits" — i.e., comparison of a vibration value to a vibration threshold value;
• the processor "accepts a weight as valid if the external vibrations do not exceed the limit at any time while the weight is being determined and otherwise initiates an alternative weight determining method" (this is the abstract, verbatim) — i.e., determining validity of weight data based on the comparison;
• the accelerometer's "predetermined limits" are adjustable by the processor "to allow processor 30 to alter the predetermined limits for various applications" — i.e., a dynamic threshold;
• aborting the weighing algorithm and initiating an alternative method (claim 12(d)), where the aborted/alternative approach is a "second weighing algorithm which is less sensitive to external vibration" (claim 13) — i.e., handling/redressing invalid weight data;
• the accelerometer is fixed to the member of the four-bar linkage that is fixed to mechanical ground (claim 10) — i.e., a vibration sensor mounted to the scale's fixed structure;
• express recognition of the settling/damping problem: scale oscillations "are damped, but only slowly arrive at a new stable output value," and abort if weight is not determined within a predetermined time of "approximately 100 to 120 milliseconds" (claim 19).

Primary reference B — Feinland et al., Pitney Bowes

US 5,172,783 A (1992-12-22). Title: "Weighing scale with validating reference channel." (Confirmed in the parent's citation list: "US5172783A (en) * 1991-07-01 1992-12-22 Pitney Bowes, Inc. Weighing scale with validating reference channel" at https://patents.google.com/patent/US10612964#8.) This reference is directed to validating a scale's weight output against a reference channel — i.e., the concept of gating a weight reading on a validity determination, in the same assignee's family as Eaton.

Secondary reference C — Huebler

US 6,907,409 B1 (2005-06-14), Francotyp-Postalia. Title: "Method and arrangement for determining a weight with a dynamic scale." (Listed in the '339 front page and in the parent's citation list.) Discloses determining weight in a dynamic (in-motion / vibration-affected) weighing environment — i.e., coping with vibration-corrupted load-cell output.

Secondary reference D — Brandorff, Intel

US 5,561,274 A (1996-10-01), Intel Corporation. Title (per parent's citation list): "Apparatus for weighing objects." (The '339 front page renders the inventor name as "Brandonff" — I am not auto-correcting that.)

Secondary reference E — Garrone, Apple

US 2012/0270605 A1 (2012-10-25). Title: "Vibration Sensing System and Method for Categorizing Portable Device Context and Modifying Device Operation." As listed on both the '339 front page and the parent's citation list. Discloses sensing vibration and using the sensed vibration to categorize a state and then modify operation — i.e., the general architecture of "measure vibration → determine a value → branch system behavior on it."

Secondary reference F — Lombriser

US 2014/0142871 A1 (2014-05-29), classified at G01M 7/022 (vibration testing). A vibration-measurement teaching.

Secondary reference G — Sawano

US 2015/0046120 A1 (2015-02-12), classified at G01G 19/50 (weighing apparatus / body weight). A weighing teaching.

Secondary reference H — retail/inventory-monitoring art (the "fixture" side of the claims)

• US 7,949,568 B2 (Fano et al., 2011-05-24); US 8,009,864 B2 (Linaker et al., 2011-08-30); US 8,189,855 B2 (Opalach et al., 2012-05-29); US 8,630,924 B2 (Groenevelt et al., 2014-01-14) — all Accenture Global Services, all directed to determination of product display parameters / inventory conditions / planogram extraction / stock-out detection based on image processing at retail fixtures.
• US 2012/0245969 A1 (Campbell, 2012-09-27) — "On-Shelf Tracking System."
• US 7,225,980 B2 (Kat et al., Industrial Technology Research Institute, 2007-06-05) — "Shopping cart with RFID capability."
• US 2013/0284806 A1 (Margalit, 2013-10-31) — "Automated purchasing system."
• US 9,235,928 B2 (Medioni et al., 2016-01-12); US 9,473,747 B2 (Kobes et al., 2016-10-18); US 9,911,290 B1 (Zalewski et al., 2018-03-06) — as listed; the OCR of the '339 list does not show titles for all of these and I have not verified them, so I do not rely on them for specific disclosures.
• US 2014/0114708 A1 (Campbell, 2014-04-24) — related on-shelf tracking disclosure.

Non-patent literature cited

• "Vibration Measuring Instruments," Quick Guide to Precision Measuring Instruments, Mitutoyo, pp. 46–47 (http://www.mitoyo.co.jp/cng/products/menu/QuickGuide_Vibration-Testing-Machines.pdf, as printed — note the OCR spells "Mintoyo"/"mitoyo"; I do not auto-correct). This establishes that vibration measurement is a textbook, off-the-shelf engineering discipline.

Same-inventor family (context, not § 103 prior art)

• US 10,520,352 B2, O'Neill & Shi, "System to validate load cell data," filed 2016-12-22, granted 2019-12-31 (https://wiki.golden.com/wiki/US_Patent_10520352_System_to_validate_load_cell_data-MARRV9G). Same inventors as '339, filed one day after the '339 priority application (2016-12-21), and so it is not "by another" and is not properly usable as § 103 prior art against '339. It is, however, strong contextual evidence that the applicants themselves regarded "validate load cell data" as a conventional, parallel line of work.
• US 10,121,121 B2, "Smart shelves" (Amazon; abstract at https://www.patents-review.com/a/14981199-smart-shelves.html): interposer assemblies between a shelf and its supports, with strain-gage load cells, processing analog signals to determine the mass of an item placed on or removed from the shelf. This is directly relevant to the "fixture + load cell → item mass → interaction" half of the claims (issued 2018-11-06, so available as prior art only as to its 2015-era priority — flag as needing date verification before reliance).

---

Combination 1 — Eaton (± Feinland) in view of the shelf/inventory art → claims 1, 2, 5–9, 11, 12, 14–19

Proposed combination: Eaton (US 2004/0009277 A1 / US 6,839,697) as the primary reference, in view of one or more of Groenevelt US 8,630,924, Fano US 7,949,568, Linaker US 8,009,864, Opalach US 8,189,855, or Campbell US 2012/0245969.

Why all elements are present:

Motivation to combine (the crux). The motivation is not speculative; it is stated in the references themselves:

1. Eaton expressly frames the problem as one of vibration-corrupted weighing and states the object as determining weight "in an environment where periods of external vibrations … can occur." That is the same problem the '339 patent recites. A PHOSITA looking for a way to make weight-based shelf monitoring robust would be drawn directly to Eaton.
2. Both references are in the same field of endeavor (load-cell weighing and weight-derived monitoring of items on a support) and are reasonably pertinent to the problem: the retail references establish the application (weight changes at a fixture = pick/place events) and Eaton supplies the robustness technique (vibration-threshold gating of weight validity). Under KSR and In re Bigio, same-field and reasonably-pertinent-art references are combinable.
3. The combination is a predictable use of known elements according to known methods: gating a measurement on an environmental-validity signal is a classic "measure the disturbance, compare to a limit, accept/reject" architecture, and Eaton already applies it to the same physical quantity (load-cell weight) subject to the same disturbance (external vibration).
4. A finite number of predictable solutions: faced with vibration-induced false pick/place events, the artisan's options are essentially (a) filter the weight signal, (b) measure and subtract the vibration, or (c) gate validity on the vibration level. Eaton's own background section enumerates exactly this set ("use sophisticated weighing algorithms … another approach … is to separately measure the external vibration and subtract it out"). Choosing option (c) is a design choice among a finite set of known alternatives — the KSR "predictable solutions" rationale.

---

Combination 2 — Eaton in view of a digital-filtering reference → claims 3, 19, 20

Proposed combination: Eaton + a weight-signal filtering teaching (Eaton's own low-pass filter 28; Huebler US 6,907,409; Sawano US 2015/0046120 A1; or the well-known notch/adaptive-filter weighing art, e.g., the Japanese publications on filter automatic-regulating functions and noise elimination that appear in the CPC G01G 23/10 "Means for damping oscillations by electric or magnetic means" class in which the '339 patent is itself classified).

Why obvious:

• Claim 3 (apply a filter to the weight data to generate filtered weight data; determine interaction data based on the filtered weight data) is squarely met: Eaton itself discloses a low-pass filter 28 in the weight-signal path ("filter 28 can be, and preferably is, implemented by software running in microprocessor 30"). Substituting or adding a filter in response to the vibration condition is the predictable use of Eaton's own filter element.
• Claim 19 (apply a filter; [use filtered weight data]) is the same technique framed as a method step.
• Claim 20 (a notch filter attenuating weight-data values representative of vibration within a frequency range, with the filtered data having lower amplitude in that range) is met by the combination, because a notch filter is the archetypal filter for attenuating a known disturbance band and is a matter of routine design choice among filters (high-pass, low-pass, notch, Butterworth — the very list the '339 spec recites). The '339 patent class G01G 23/10 ("damping oscillations by electric or magnetic means") is itself populated with electric-filter-based damping art; selecting a notch characteristic to attenuate the vibration band requires no more than routine experimentation for a PHOSITA who already knows, from Eaton and from accelerometer data, the frequency content of the disturbance.
• Motivation: attenuating rather than discarding data is an obvious alternative where discarding would lose too much useful information — the '339 specification itself presents filtering and invalidation as two interchangeable alternatives ("In a second implementation, weight data … designated as invalid may be filtered"). KSR treats such a choice between two known techniques for the same purpose as obvious.

---

Combination 3 — Eaton in view of a damping/settling-time teaching → claims 4, 8, 13

Proposed combination: Eaton + the settling-time/damping teachings (Eaton's own recognition that oscillations "are damped, but only slowly arrive at a new stable output value," and its predetermined abort window of "approximately 100 to 120 milliseconds"; augmented by the mechanical-damping art cited in the same CPC class, e.g., anti-vibration damper/dummy-load-cell teachings such as EP 0 129 249 B1 and EP 0 511 217 B1/US 5,347,092, which quantify how vibration energy dissipates and how long a weighing signal takes to settle).

Why obvious:

• Claim 4 (determine a settling interval; determine a second interval after the first equivalent to the settling interval; designate weight data of the first and second intervals as invalid) is a direct and predictable extension of Eaton: Eaton already aborts and waits, and already deals in a time-bounded response window (100–120 ms). Extending the invalid region past the vibration event by a settling period is the purely conventional engineering step of accounting for the mechanical system's decay time — a fact of second-order underdamped systems that Eaton itself recites ("because systems such as scale system 10 are underdamped 2nd order mechanical systems the time required for the disturbances … to decay will vary proportionately with the magnitude of the disturbances").
• Claim 8 (second independent system: designate first-interval and second-interval weight data as invalid) reaches the same result structurally and is met by the same teachings.
• Claim 13 (settling interval based on construction of the structure, previously determined total weight on the structure, or placement of the vibration sensor): each enumerated basis is a known determinant of a mechanical system's settling time (mass, stiffness, sensor location). Deriving a settling interval from a loaded mass and a mounting location is routine optimization of a result-effective variable, and the '339 specification concedes as much: "the settling interval … may be varied with respect to the total weight previously measured on the shelf."

---

Anticipation note (not § 103, but relevant to the overall analysis)

Because Eaton's disclosure is so close, independent claim 1 may be vulnerable to a § 102 anticipation attack under Eaton alone, if "structure" is given its broadest reasonable construction. Eaton discloses the complete control loop: load cell → weight data; accelerometer → vibration data; vibration value; comparison to a predetermined limit; and a validity determination ("accepts a weight as valid"). The only potential gap is the word "structure" (Eaton's is a postage-scale platform/load-cell frame, not a retail shelf). Given that claim 1 recites no interaction data and no retail context, that gap is narrow. If the examiner/defendant reads "structure" broadly (which the specification permits — the spec states "the techniques described herein may be used with regard to other types of fixtures 102 and devices that may stow items, such as racks, hangers, bins, and so forth"), Eaton alone anticipates claim 1 and renders claims 2, 5, 6, 9, 11, 12, 14–19 obvious. I flag this as claim-1 scope risk rather than asserting it as a conclusion, because I have not read Eaton's complete specification line-by-line against every claim term.

---

Claim-by-claim § 103 conclusion

Net assessment: Every independent claim (1, 8, 15) and every dependent claim is, on this record, obvious under § 103.

---

The patentee's best (and, in my view, insufficient) rebuttals

A competent patentee would argue as follows, and here is my assessment of each:

1. "Eaton is a postage scale; the '339 patent is a retail shelf-inventory system. Different fields."
   Weak. Under KSR and In re Bigio, art is analogous if it is in the same field of endeavor or reasonably pertinent to the problem. Both address load-cell weighing corrupted by external vibration, and the '339 patent's own stated problem — "a large truck passing by the facility may generate vibrations that produce erroneous output" — is exactly Eaton's "external (e.g. ground) vibrations." A PHOSITA solving shelf-vibration false events is on notice of the postal-scale art.

2. "Eaton aborts and re-weighs; it does not 'designate the weight data as invalid' or 'disregard' it the way claim 2 requires."
   Weak. Eaton's abstract and claim 12(c)–(d) expressly refuse to accept a weight obtained during excess vibration and instead follow an alternative route. Labelling that "invalid" and "disregarding" it is semantic, not technical; the claims' "disregard" is met by any use of a validity flag to exclude the reading from the downstream decision (and dependent claim 3/19's filtering alternative shows the applicant treated "discard vs. filter" as interchangeable).

3. "The claims require contemporaneous acquisition of weight and vibration data during the same first interval; Eaton's accelerometer is checked 'at any time during' the determination, which is different." (This is the strongest form of the argument, and it is the point the earlier "Patent summary" also flagged.)
   Weak-to-moderate, and it only touches claim 12 (dependent: "at least a portion … contemporaneous"). Eaton's "at any time while the weight is being determined" is, if anything, broader than "during the first interval" and plainly encompasses simultaneous acquisition. Moreover, claims 1, 8, and 15 do not require more than acquisition "during the first interval."

4. "The dynamic threshold of claims 13/17 based on structure construction, total weight, sensor placement, and rigidity is a specific, non-obvious parameterization."
   Weak. Eaton expressly teaches that the limits are alterable by the processor "for various applications," and the enumerated bases are the known physical determinants of a loaded structure's response to vibration. Choosing a threshold as a function of load and mounting is routine optimization of a result-effective variable (In re Applied Materials).

5. "No motivation to combine the retail references with Eaton."
   Weak. The problem to be solved supplies the motivation: retail weight-based shelf monitoring was known (Groenevelt, Fano, Linaker, Opalach, Campbell, and Amazon's own "Smart shelves" US 10,121,121), and the known defect of weight-based shelf monitoring is false events from vibration — the very defect the '339 patent opens with. When the prior art discloses the problem and the prior art discloses a solution to that problem in the same physical domain, the combination is obvious.

---

Objective indicia (§ 103 factor 4)

• No litigation or PTAB proceeding involving the '339 patent was located, per the previously generated litigation section — so there is currently no adjudicative record of validity challenges, settlements, or injunctions that would bear on objective indicia.
• No evidence of unexpected results, long-felt but unmet need, failure of others, or industry praise is present on the face of the patent or in the retrieved record. The specification asserts only incremental efficiency and accuracy benefits ("noise … may be minimized," "fewer resources are needed," "fewer instances of interaction data with low confidence values occur") — these are the predictable consequences of the claimed technique and are not the kind of "unexpected results" that rebut a prima facie case. Under KSR/Pfizer, an assertion of generic improvement without demonstrated surprise carries little weight.
• Any commercial-success argument would face a weak nexus: Amazon's shelf-inventory deployments are driven by many technologies (image sensors, RFID, light curtains, instrumented auto-facing units — all recited in the '339 specification itself), not by the vibration-gating feature alone. Without a nexus, commercial success is not probative.
• Copying/simultaneous invention would if anything cut against patentability: the same inventor pair pursued a parallel "validate load cell data" application (US 10,520,352) filed one day later, and multiple Accenture/independent references were already addressing fixture-based inventory monitoring — evidence that this was a crowded, incremental field rather than a breakthrough.

---

Bottom line

Using the prior art of record on the face of US 10,948,339 (item 56) — principally Eaton et al., US 2004/0009277 A1 / US 6,839,697 B2 (Pitney Bowes), "System and method for weighing items such as mailpieces in the presence of external vibration," in combination with Feinland, US 5,172,783, Huebler, US 6,907,409, and the retail shelf-monitoring art (Groenevelt US 8,630,924; Fano US 7,949,568; Linaker US 8,009,864; Opalach US 8,189,855; Campbell US 2012/0245969):

• Independent claims 1, 8, and 15 are obvious under § 103, and claim 1 is additionally at material risk of § 102 anticipation by Eaton alone.
• All dependent claims 2–7, 9–14, and 16–20 are obvious, being either expressly taught by Eaton (validity gating on a vibration threshold; abort-and-alternative handling; low-pass filtering; adjustable limits; 100–120 ms response windows) or conventional engineering choices (RMS statistics, multi-load-cell topology, notch filtering, settling-interval parameterization by load and mounting).
• The motivation to combine is strong and articulated by the references themselves: Eaton states the precise problem ("external (e.g. ground) vibrations … can slow or even prevent the scale system output from converging"), discloses the precise solution (accelerometer + limit detector + validity gating), and the retail references supply the shelf/fixture application that animates the '339 claims. This is precisely the KSR scenario of a predictable use of prior-art elements to solve a known problem with a finite number of predictable solutions.

Two honesty flags to carry forward: (1) I could not retrieve the foreign patent document column of the (56) list, so additional foreign art may exist that would only strengthen (never weaken) an obviousness position; and (2) the OCR of the (56) list is partially garbled (e.g., 9/2016 4:41 B1 3/2018 Worley et al.), so those specific entries should be verified against the USPTO/Google Patents image before being cited in any formal paper.