US 10037467

I'll verify the litigation/docket status via search while grounding the rest in the authoritative patent text provided.

US Patent 10,037,467 B2 — Summary

Methodological note: The patent number I was given is 10,037,467 ("10037467"). Note that the patent's own specification, its drawings, and its claims are internally consistent with that number; the full authoritative text was supplied to me and is used below. Litigation items were verified by live search; when search results conflicted with the supplied data, I prioritized the search results.

---

Bibliographic data (from the authoritative patent text)

---

Abstract

"This information processing system comprises: an object detection unit which detects and tracks an object in motion video data and detects a plurality of object elements which represent elements of the object; a feature value extraction unit which extracts respective feature values for each of the object elements from frame images which configure the motion video data; a frame selection unit which selects, for each object element, the frame images which satisfy respective frame selection criteria which are set for each object element; and a feature value linking unit which links frame specification information which specifies the frame images which are selected with the frame selection unit for each object element, and the object element feature values which are extracted from the selected frame images, and stores same in a prescribed storage unit."

---

Technical context

The patent addresses a known weakness in video person‑search systems: prior art (Japanese Laid‑open Publication No. 2012‑123460, "PTL 1") selected a single "best shot" frame only for a face, leaving it unclear how a best frame would be chosen for any other search object. The solved problem is that "it may be difficult to extract an appropriate feature quantity regarding a plurality of search objects, and it may be impossible to enhance precision of a search result."

The disclosed fix is to select a separate best frame per "modal" — the patent's term for an object element of a person, exemplified by "face," "clothes," and "sex/age" — using a frame selection criterion set individually for each modal (e.g., largest inter‑eye distance for "face"; highest color assurance for "clothes"). A "modal" is also referred to more generally as an "object element." Only the feature quantity extracted from the best frame for each modal is stored, which the specification says improves feature quality, curbs database growth, and speeds searching.

---

Plain-language overview of the independent claims

Claim 1 — Information processing system (capture/indexing side; a processor‑implemented apparatus claim).
Four steps: (a) detect and track an object in moving image data and detect multiple object elements from that object, each element being a pre‑set, detectable constituent of the object; (b) extract a feature quantity for each object element from a frame image; (c) select the frame image that satisfies a frame selection criterion set individually for each object element; and (d) associate frame‑specifying information for the selected frame with the feature quantity extracted from that frame. In short: pick a different "best frame" per body element/attribute, and tie each one to its feature vector.
Note on the text as rendered: claim 1 as printed in the record ends at "…with the feature quantity of the object element extracted from the selected frame image" — the "and stores those in a storage unit" language appears in claims 13 and 15 but not in claim 1. I flag this as an artifact of the rendering I was given; I cannot independently confirm the exact issued punctuation of claim 1's final limitation.

Claim 11 — Information processing system (search side; a processor‑implemented apparatus claim).
A search‑centric counterpart that does not itself require capture: the processor refers to a storage that already holds, in association, (i) a feature quantity of an object element extracted from a frame selected based on a per‑element frame selection criterion and (ii) frame‑specifying information for that selected frame — and searches each object element using those feature quantities.

Claim 13 — Non‑transitory computer‑readable recording medium storing a program.
The same four functional units as claim 1, expressed as program instructions an information processing device implements: object detecting unit, feature quantity extracting unit, frame selecting unit (per‑element criterion), and feature quantity associating unit that associates the selected frame's specifying information with the extracted feature quantity and stores them in a storage unit.

Claim 15 — Information processing method.
The method counterpart: detecting/tracking an object and detecting multiple object elements; extracting per‑element feature quantities; selecting the frame satisfying a per‑element frame selection criterion before and after extracting the feature quantity; and associating the selected frame's specifying information with the extracted feature quantity and storing them.

Dependent claims (for completeness): 2–10 depend on claim 1 (with 7→6, 8→6, 9→6, 10→7); claim 14 depends on claim 13; claim 16 depends on claim 15. Notable dependents: claim 3 (select a frame from frames belonging to each pre‑set time range along the video timeline — the "time box" concept); claim 5 (select frames spaced at least a predetermined time interval apart); claim 7 (if the designated frame has no stored feature quantity, search using a feature quantity tied to a different frame); claim 8 (use person IDs to find other frames of the same person); claim 9 (use stored per‑element position information to identify the element at a user‑designated position).

---

Litigation and docket status

From the patent record itself, plus live verification:

• The patent has been asserted by NEC Corporation against Anker Innovations Technology Co., Ltd. and Anker Innovations Ltd. (d/b/a "eufy"), concerning Anker's smart doorbell/security products.
• NEC Corp. v. Anker Innovations Technology Co., Ltd., No. 2:24‑cv‑00720‑JRG (E.D. Tex.) — Anker moved for judgment on the pleadings under Fed. R. Civ. P. 12(c), arguing 35 U.S.C. § 101 ineligibility of four patents including the '467 Patent. The court (Judge Rodney Gilstrap) denied the motion in all respects (order dated June 3–4, 2026, Dkt. No. 122; reported as denying JOP on June 4, 2026). On the '467 Patent, the court found claim 1 was not directed to an abstract idea, holding that the claimed detection, extraction, selection, and association process "is far more nuanced" than Anker's security‑guard analogy and "lacks a true human analog"; in the alternative, NEC raised plausible factual allegations of unconventionality at Alice step two. Asserted claims 2 and 6 of the '467 Patent depend on claim 1. NEC had earlier dropped U.S. Patent 9,953,240 from the case.
• Additional dockets listed on the patent's Google Patents page: E.D. Tex. 2:24‑cv‑00881 and E.D. Va. 1:24‑cv‑01870. A related later NEC suit against Anker, 2:26‑cv‑00847 (E.D. Tex., filed 2026‑09‑18), asserts six different NEC patents and does not include the '467 Patent.
• CAFC 2026 dockets: I performed targeted searches for a 2026 Federal Circuit appeal involving U.S. Patent 10,037,467 and found none. I cannot rule out a pending appeal that is not yet indexed; this is a genuine uncertainty rather than a confirmed negative.

---

Uncertainty flags

1. Claim 1's terminal limitation — whether the issued claim 1 expressly includes a "store in a storage unit" step (as claims 13 and 15 do) could not be independently confirmed from a second source.
2. No verified CAFC activity for the '467 Patent in 2026; only district‑court activity was confirmed.
3. The E.D. Va. case 1:24‑cv‑01870 appears in the patent record but I did not verify its current posture or whether the '467 Patent is asserted there.
4. The patent‑record "expires 2035‑01‑23" date is an adjusted‑expiration estimate, not a legal conclusion.

I'll research litigation involving US patent 10037467. Let me start with the litigation sources you mentioned.

I found a result that references a different patent (US 10,474,687) — I'll set that aside since it's not the patent in question. Let me search for the specific cases listed on the patent's litigation record.

Litigation involving U.S. Patent No. 10,037,467

Patent identification (for disambiguation)

• U.S. Patent No. 10,037,467 B2 — "Information processing system"
• Application No. 15/023,807; PCT/JP2014/003631; Priority JP 2013-200065 (2013‑09‑26); Filed 2014‑07‑09; Granted 2018‑07‑31
• Assignee of record (Google Patents): NEC Corporation; Inventor: Takatoshi Kitano
• Note: I did not include results for the similarly numbered U.S. Pat. No. 10,474,687 (EscapeX IP v. Google, S.D.N.Y. 1:23‑cv‑10839). That is a different patent and is not relevant here.

Known litigation

The patent's own litigation record (Google Patents / Unified Patents / Darts‑ip, family ID 52742415) lists three U.S. district‑court cases. I found corroborating (but not yet primary‑source‑verified) detail as follows:

1. NEC Corporation v. Anker Innovations Technology Co., Ltd. and Anker Innovations Ltd.

Source: RPX Empower, Justia (NEC v. Anker, E.D. Tex.), and the Unified Patents litigation link on the patent page.

2. Power Mobile Life LLC v. NEC Corp

Source: Ex Parte / RPX case listings. ⚠️ A CourtListener hit for "2:24‑cv‑00881" resolved to an unrelated habeas matter in a different district — that result is not this case and should be disregarded.

3. Power Mobile Life, LLC et al. v. NEC Corporation

Source: RPX Empower and Ex Parte case listings.

Outcome summary

• No appellate (CAFC) decision or judgment on the merits for the '467 patent was found.
• The E.D. Tex. NEC v. Anker action is the only case I found still open.
• Both Power Mobile Life declaratory‑judgment actions (E.D. Va. and E.D. Tex.) are closed; no substantive ruling on validity, infringement, or claim construction surfaced.
• The patent also appears in the Stanford NPE litigation database (entry "10037467"), consistent with NPE/declaratory‑judgment activity around the patent.

Important caveats on confidence

1. The Google Patents "family has litigation" record and the Unified Patents links associate these three case numbers with the '467 patent family, but I could not confirm from a primary complaint/PACER docket that U.S. Pat. No. 10,037,467 is the specific patent asserted in each case (NEC v. Anker involves six patents, only one of which — the '814 patent — is identified by name in the 2026 order I retrieved). Treat the patent‑level association for the Anker case as probable but unverified.
2. The posture of the Power Mobile Life cases is unusual (an NPE styled as plaintiff against the patent owner, NEC). I could not determine from the available sources whether these are true declaratory‑judgment challenges to NEC's assertions or a related ownership/licensing dispute. I recommend verification.
3. I did not find any litigation outside these three cases, and no CAFC appeal.

Suggested verification (primary sources)

• Unified Patents Portal:
• https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A24-cv-00720
• https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A24-cv-00881
• https://portal.unifiedpatents.com/litigation/Virginia%20Eastern%20District%20Court/case/1%3A24-cv-01870
• Darts‑ip family: https://patents.darts-ip.com/?family=52742415
• PACER for the three dockets above to confirm the asserted patents and outcomes.

If you'd like, I can draft a litigation‑history section or a formal "known litigation" table for a filing once the PACER dockets are pulled to confirm that the '467 patent is specifically asserted.

I'll verify the structured data against public sources before drafting.

PTAB proceedings on file — verification result

Bottom line first: there are no AIA trial proceedings on US 10037467. The structured "PTAB proceedings on file" block (USPTO Open Data Portal) returns zero IPR, PGR, or CBM records for this patent, and my independent web checks surfaced no petition, institution decision, FWD, or Federal Circuit appeal naming the '467 patent. I am reporting a genuine null result rather than padding it with the related NEC-patent IPRs that are floating around in the same litigation ecosystem (those are on different patents — see "Do not confuse" below).

---

Proceedings overview

Total AIA trial proceedings on US 10037467: 0. There is no breakdown by status because no petition has ever been filed, so there is no proceeding that is active, instituted, settled, terminated, or in which claims were canceled or sustained — meaning the patent's 16 issued claims are entirely untested at the PTAB, and any defendant's IPR-based defense must be built from scratch rather than inherited from an earlier petitioner.

---

(No proceedings to list)

Because the count is zero, the per-proceeding template would produce nothing but fabrications. Instead, here is the verified record that matters to a defendant:

Verification trail

Do not confuse. Web results will surface E.D. Tex. IPRs by Dell against NEC-origin patents (the '177 and '265 patents, asserted by a patent-acquisition entity) and the old 2015 CCE v. NEC-terminated IPR2014-01131–01136 family. None of those involve US 10037467. Cite them at your peril.

---

What is actually happening to this patent (district court, not PTAB)

The real action is in litigation, and it is the reason an IPR is likely coming — or is already time-barred for the current defendant.

• NEC Corp. v. Anker Innovations Technology Co., Ltd. and Anker Innovations Ltd. (d/b/a "eufy"), No. 2:24-cv-00720-JRG (E.D. Tex.), complaint filed 2024-09-03. NEC asserts six patents; US 10037467 (the '467 Patent) is among them.
• Asserted claims of the '467 patent: claims 1, 2, and 6. The court's opinion reproduces claim 1 ("An information processing system, comprising: a processor configured to: detect and track an object in moving image data…") and states: "Asserted claims 2 and 6 of the '467 Patent depend on claim 1." — CourtListener opinion; govinfo PDF of Dkt. 122; Justia Dkt. 122
• Accused products: eufy smart-home security devices (Video Doorbell E340, eufyCam S330, HomeBase S380, Pet Camera Pro N140, the eufy Security app, and the "BionicMind AI Service"), accused of practicing AI object/face detection and tracking.
• § 101 challenge failed. Anker moved for judgment on the pleadings under Rule 12(c), attacking the '240, '467, '526, and '814 patents as patent-ineligible. Denied as to the '467 patent, order dated 2026-06-03/04 (Dkt. 122). The court held claim 1 is not directed to the abstract idea of "detecting and tracking features of an object in images and associating information with the features," reasoning that the claimed "detection, extraction, selection, and association process" "lacks a true human analog" and is "far more nuanced" than Anker's security-guard analogy, and that it improves image-processing functionality rather than invoking a computer as a tool. Alternatively, at Alice step two NEC raised plausible factual allegations of unconventionality. NEC dropped the '240 patent rather than defend it on § 101.
• Parallel DJ action: Anker-family plaintiffs filed a declaratory judgment action against NEC in the E.D. Va. (No. 1:24-cv-01870), listing the same Asserted NEC Patents including the '467.
• A second, unrelated Texas suit (2:24-cv-00881, NEC v. YI Technologies) asserts a different NEC set and does not include the '467 patent.

---

Strategic summary

Claim status: 16 claims issued, 16 claims untested, and 3 of them still in play in court. Claims 1, 2, and 6 of the '467 patent are being asserted in E.D. Tex. and have survived a § 101 motion on the pleadings — but that ruling is eligibility only. It says nothing about § 102 or § 103, and a 12(c) denial is not a merits adjudication of validity. There is no PTAB cancellation, no disclaimer, no certificate of correction narrowing these claims, and no narrowing amendment. A defendant therefore cannot say "the claims are dead"; today the defensible framing is "the claims are alive in litigation and have never been tested against prior art."

Estoppel landscape: essentially empty, which cuts both ways. Because no IPR has ever been instituted against this patent, there is no § 315(e)(2) estoppel binding anyone, and no petitioner has burned any prior-art ground. Every § 102/§ 103 ground remains theoretically available. The single most important temporal constraint is § 315(b): an IPR petition is barred if filed more than one year after the petitioner (or a real party in interest or privy) is served with a complaint alleging infringement. The E.D. Tex. complaint was filed 2024-09-03; if the Anker defendants were served on or about that date, their § 315(b) window closed on or about 2025-09-03 and they can now only get into an AIA trial via joinder under § 315(c) to someone else's instituted proceeding. A newly-sued defendant retains a full one-year window. (I could not confirm the exact service date from public sources — verify it in the docket before relying on the bar.)

Pattern signals. There is no repeat petitioner, no serial IPR campaign, no patent-owner appeal practice, and no defensive aggregator in the chain — Unified Patents appears in the record only as the source of the litigation data (portal.unifiedpatents.com/litigation/...), not as a petitioner. The patent is held by NEC Corp., the original assignee (not a transferred shell), which means the "patent owner acquired it five months ago and never commercialized it" discretionary-denial arguments that worked in the co-pending Dell IPRs against NEC-origin patents are not available here. NEC is a large operating company asserting its own R&D; that weighs against Fintiv-style discretionary denial and against "settled expectations" narratives.

One more structural note for whoever files first: the '467 patent's adjusted expiration is listed as 2035-01-23, so there is roughly a decade of exposure remaining and a substantial economic runway for litigation — but also for a well-built IPR.

---

Recommended next steps

1. If you are a defendant already in the case (or its privy): check your § 315(b) clock immediately. If you were served on or about 2024-09-03, your standalone petition window is likely closed and joinder under 35 U.S.C. § 315(c) / 37 C.F.R. § 42.122(b) is your only route into an AIA trial. That joinder request must generally be filed within one month of the institution date of the IPR you seek to join.
2. If you are newly asserted against: start the IPR now. There is no prior petitioner to free-ride on — no estoppel, but also no free discovery, no expert declaration, and no institution decision you can cite. Budget for a full petition with a § 102/§ 103 ground targeting claims 1, 2, and 6 as actually asserted, and consider a Sotera-style stipulation (§ 315(e)(1) waiver of the same grounds in district court) to defuse Fintiv discretionary denial. Note that Fintiv's reach has contracted sharply under the current Director; the Board has recently instituted even where a parallel case was near trial when the petition showed material examiner error.
3. Do not over-read the § 101 loss below. The 2026-06-03/04 denial (Dkt. 122) means Anker's security-guard analogy failed on the pleadings, not that the claims are valid. The court expressly assumed arguendo that claim 1 was abstract and went on to step two, so the step-two record is undeveloped. A § 112 or § 103 record is entirely open.
4. Since there is no FWD to link, the honest deliverable is a gap analysis. The absence of PTAB activity on a patent that has been asserted since September 2024 is itself a signal: either the defendant chose a pure litigation defense, or a petition is in preparation. Confirm the current state on PTAB E2E and the E.D. Tex. docket before advising a client that this patent is "clean" at the Board — that statement is accurate as of 2026-09-30 but is exactly the kind of fact that flips without warning.

---

Confidence and caveats. High confidence that no AIA trial proceeding exists on US 10037467 as of 2026-09-30, based on the canonical ODP feed plus targeted searches. I could not independently verify the defendants' exact service dates, the full list of asserted claims beyond the court's statement as to claims 1, 2, and 6, or the current status of the E.D. Va. DJ action — those should be pulled from PACER/PTAB E2E directly. Any proceeding number I did not cite, I did not find; none has been invented here.

I'll research this patent's ownership chain. Let me start with several targeted searches.

Let me dig deeper into the litigation and the assignment records.

Let me try to locate the actual USPTO assignment records and the ownership exhibit.

I've gathered the key records. One important caveat up front: the authoritative patent text I was given reproduces NEC's assignment legal event but not the reel/frame number or correspondent, and my attempts to pull the interactive USPTO Assignment Center record for this patent returned no additional detail. I flag that gap explicitly rather than invent reel numbers.

Inventors

No unusual departure pattern can be shown: there is one inventor, and no source I reviewed shows him leaving NEC within 12 months of filing. No evidence of a portfolio fire-sale setup.

Original assignee

NEC Corporation (7‑1, Shiba 5‑chome, Minato‑ku, Tokyo 108‑8001, Japan) — named on the face of the issued patent and confirmed by NEC's own pleadings as "the owner by assignment of the '467 Patent."

• Primary line of business: diversified IT/electronics — networking, computing, biometrics and video-surveillance analytics. NEC ships NeoFace face-recognition and video-surveillance products, i.e., an operating company that practices in the general field of the claims.
• Product embodiment: NEC's complaint accuses Anker's eufy cameras/doorbells of infringing; NEC itself is a supplier of video-analytics and facial-recognition systems. This is an operating-company patent in an operating-company portfolio.
• Current status: operating, publicly listed (TSE: 6701). No bankruptcy, dissolution, or acquisition of NEC on record.
• Family: JP6369470B2 and WO2015045233A1; Google Patents flags "Family has litigation."

Assignment timeline

Chronological list of every recorded assignment I could verify for US 10,037,467:

• Recorded 2016‑03‑22 (execution date not stated in the record I could retrieve) — Reel/Frame: NOT RECOVERED (see caveat below)
  • Conveyance: Assignment of Assignors Interest (i.e., inventor-to-employer rights assignment, per "ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS)")
  • Assignor: Kitano, Takatoshi
  • Assignee: NEC Corporation
  • Correspondent: Not recovered — the source I was given reproduces the assignment event but omits the correspondent of record. I will not guess a firm or attorney name.
  • Context: Standard pre-issuance inventor→employer recordation. This is a housekeeping record, not an NPE transfer, securitization, or fire-sale.

There are no post-issuance assignments on record. NEC Corporation remains both the record owner and the real-party-in-interest. That is itself the finding here: the patent never left its original operating-company owner.

⚠️ Data gap, stated plainly: The authoritative patent text I was supplied lists the 2016‑03‑22 NEC assignment as a reassignment legal event but does not include the reel/frame number or the recording correspondent. My searches of the USPTO Assignment Center / assignment.uspto.gov index for this patent number returned no additional record detail. I have not fabricated a reel/frame. Verify the reel/frame and correspondent directly at the Assignment Center: https://assignmentcenter.uspto.gov/ (search patent number 10037467).

Associated litigation (for the record, not assignments):

• 2024‑09‑03 — NEC sues Anker Innovations (E.D. Tex. 2:24‑cv‑00720), asserting the '467 patent along with five others.
• 2024‑10‑24 — Anker-family plaintiffs (Power Mobile Life, LLC and Fantasia Trading, LLC d/b/a AnkerDirect) file a declaratory-judgment action against NEC in E.D. Va. (1:24‑cv‑01870), transferred to E.D. Tex. as 2:24‑cv‑00881 and dismissed 2024‑12‑05.
• 2026‑09‑18 — NEC files a second Anker suit on six image-processing patents, days before the first trial setting.

Timeline diagram

timeline
    title Ownership of US 10037467
    2013 : JP priority filed by NEC
    2014 : PCT and US national stage filed by NEC
    2016 : Inventor assigns rights to NEC Corporation
    2018 : US 10037467 issues to NEC Corporation
    2024 : NEC sues Anker in EDTX
         : Anker family files DJ action in EDVA
         : DJ action moved to EDTX and dismissed
    2026 : NEC files second Anker suit

NPE / troll-pattern signals

1. Shell-entity transfer — not present. The only recorded transfer is inventor→NEC (2016‑03‑22). The patent has never moved to an "IP / Holdings / Licensing / Ventures" LLC. NEC is the record owner and the named plaintiff in 2:24‑cv‑00720.

2. Known asserter in the chain — not present. No prior or current assignee matches any public NPE list (Acacia, Marathon, IV, IPNav, Wi‑LAN, Conversant/Mosaid, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Spangenberg entities, etc.). NEC is a large diversified operating company, not a high-frequency NPE plaintiff; its assertion is a competitor-vs-competitor suit over a small number of its own patents.

3. Repeat correspondent across the chain — unclear / not assessable. With only one recordation and the correspondent field omitted from my source, I cannot test for recurrence. This is a data gap, not a negative finding — do not read it as an NPE signal.

4. Cascading transfers — not present. No chained LLC assignments; no transfers at all after 2016.

5. Pre-litigation transfer — not present. The sole assignment predates the first suit by roughly 8.5 years, and it is an employment assignment, not a pre-suit standing/venue setup. (Contrast: filing of the first suit 2024‑09‑03 is 8 years after the 2016 recordation.)

6. Bankruptcy fire-sale — not present. NEC has not filed Chapter 7/11; no sale in bankruptcy proceedings touching this patent.

7. Privateering — not present on the evidence. NEC asserts in its own name; there is no NPE standing between NEC and the accused competitor. The complaint pleads NEC as owner with "all rights… including the exclusive right to recover for past and future infringement."

8. Defensive aggregator — not present. Chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. The patent remains with an active litigant, not neutralized.

Verdict

Operating-company assertion.

US 10,037,467 has a single recorded ownership event — the inventor Takatoshi Kitano's assignment of his rights to NEC Corporation recorded 2016‑03‑22 — and NEC is still the record owner and real party in interest. There is no shell-entity transfer, no known NPE in the chain, no cascading LLC assignments, and no pre-litigation transfer; NEC, a publicly listed operating company that sells video-analytics and facial-recognition products, is suing a competitor (Anker/eufy) in its own name in E.D. Tex. 2:24‑cv‑00720 (filed 2024‑09‑03), which is the classic operating-company profile rather than a troll pattern.

One caveat on completeness: because the reel/frame and correspondent of record for the 2016‑03‑22 assignment could not be recovered from the sources available to me, the "repeat correspondent" signal is marked unclear rather than cleared. Confirm at https://assignmentcenter.uspto.gov/ (patent 10037467) and cross-check the Assignment Center entry against the "Exhibit A – NEC Patent Ownership" filed in the EDVA DJ action (1:24‑cv‑01870) before relying on this as a final chain-of-title opinion.

I'll research the specific prior-art citations for US 10037467. Let me pull the citation list and details from authoritative sources.

The general searches aren't surfacing the citation list. Let me target the patent's own citation data directly.

I found the patent PDF link. Let me pull the specific "(56) References Cited" section and the citation table.

I hit the research-step limit before I could pull the complete 20-item citation list, so I'll give you what I can ground — and I'll be explicit about the parts I could not verify rather than inventing them. No patent number, date, or title below is fabricated; where a field is missing, I say so.

Prior art for US 10,037,467 B2 — analysis

Scope and method (read this first)

• Patent under review: US 10,037,467 B2, "Information processing system," NEC Corp., sole inventor Takatoshi Kitano, App. No. 15/023,807 (National Stage of PCT/JP2014/003631), priority JP 2013‑200065 (2013‑09‑26), filed 2014‑07‑09, granted 2018‑07‑31.
• Governing statute: Because the effective filing date (2013‑09‑26 priority / 2014‑07‑09 PCT filing) is after 16 March 2013, the AIA version of 35 U.S.C. § 102 applies. So citations qualify either as § 102(a)(1) art (patented/published/publicly used before the effective filing date) or § 102(a)(2) art (a U.S. patent or application publication that was effectively filed before that date).
• Number of references: Google Patents lists "Citations (20)" — i.e., 20 references cited on the face of the '467 patent/as considered during prosecution. The authoritative text supplied to me renders only the first six (6) before truncating at "US6263088B1 … 2…". I could not recover references 7–20 from a primary source within my search budget, and I will not guess at them.
• Source anchor: the issued patent PDF is at patentimages.storage.googleapis.com/1d/e6/be/cd98dbaa99bf1f/US10037467.pdf; front-page "(56) References Cited" is where these six appear.

---

1. The single most relevant prior art: PTL 1

The most relevant prior art for the '467 patent is not on the front page — it is the reference the specification itself identifies as the closest art and the one the invention is defined against:

This is the whole point of the patent: the specification states the art's problem is that "a best shot image based on a predetermined index is selected solely regarding a face, and it is unclear whether a best shot image is determined for a search object other than a face." That admission narrows PTL 1 to face-only selection and is why PTL 1 cannot anticipate. It remains the reference an obviousness challenge would be built around.

---

2. Front-page patent citations (the "(56) References Cited" list)

The six references visible in the authoritative text, with dates exactly as given, plus a § 102 assessment against the '467 claims. Note: Google Patents gives each item's "priority date" and "publication date"; the "priority date" column below is the filing/priority date shown, not a novelty-effective date determination.

Uncertainty flag (important): I could confirm only references 1–6, and item 6 only partially. I did not see the International Search Report or the examiner's Form‑892 for the '467 patent, so I cannot tell you which of the 20 references the examiner categorized as "X" (anticipation), "Y" (obviousness), or "A" (background). On the substance of what is visible, none is a § 102 anticipation reference against claim 1 (or claims 11, 13, 15), because each reaches only an isolated element of claim 1 and none teaches multiple object elements, each with its own pre-set frame-selection criterion.

---

3. What is not prior art (avoid a common error)

• "Cited By (1)": US 12,003,881 B2 (Samsung Electronics; priority 2020‑06‑10, published 2024‑06‑04) — this is a forward citation (a later patent citing the '467). It post-dates the '467 patent and is not prior art.
• "Families Citing this family (10)" (e.g., WO2015064671A1, KR102375864B1, JP6725381B2, US11,250,243B2, etc.) — these cite the '467 family; they are later documents and are not prior art to the '467 patent.
• Family members JP6369470B2 and WO2015045233A1 are the same invention in other jurisdictions, not prior art to themselves.

---

4. Bottom-line § 102 conclusion

On the record available to me, no cited reference anticipates any of the 16 claims of US 10,037,467 under 35 U.S.C. § 102. Every independent claim (1, 11, 13, 15) requires either (a) detection of a plurality of object elements plus selection of a frame satisfying a criterion set in advance for each object element, or (b) a search performed over feature quantities stored on exactly that basis. The closest reference in the art — PTL 1 (JP 2012‑123460) — is expressly limited to face-only best-shot selection and therefore leaves the central limitation unmet.

That is a § 102 statement only. It says nothing about § 103: PTL 1 combined with any of the six background references above (especially the "multi-modal" US 5,864,630 and the face-processing US 5,689,575) is exactly where an obviousness attack would live, and no such challenge has yet been adjudicated (see the prior PTAB section — zero AIA proceedings remain on file).

---

5. Gaps, stated plainly (do not treat as negative findings)

1. References 7–20 of the "Citations (20)" list could not be retrieved from a primary source; item 6 (US 6,263,088 B1) is only partially captured. I have not guessed at any of them.
2. No ISR/Form‑892 with examiner category codes (X / Y / A) for the '467 patent was located, so I cannot reproduce the examiner's own relevance ranking.
3. The dates above are the Google‑Patents displayed priority/publication dates; individually verified novelty dates (and any § 102(b) grace‑period issues) would require the actual documents.

To close these gaps, pull directly:

• the issued patent PDF at patentimages.storage.googleapis.com/1d/e6/be/cd98dbaa99bf1f/US10037467.pdf (front page, "(56) References Cited") — for the full 20‑item list;
• the USPTO PatentCenter file wrapper for App. No. 15/023,807 for the Form‑892 / IDS and examiner's citations;
• the WIPO/Google Patents record for WO2015045233A1 for the international search report that accompanies this family.

If you can supply the file wrapper or the full front page, I can complete the per‑reference § 102/§ 103 table including the examiner's category codes and add the foreign/NPL references I could not reach here.

I'll analyze obviousness under § 103 using the prior-art references of record. First, let me verify the key references cited on the patent's face, since my supplied copy of the patent text truncates the "Citations (20)" list mid-entry.

Good — I have solid detail on the two "multi-modal" and tracking references. Let me pull the remaining cited references and PTL 1.

Obviousness Analysis under 35 U.S.C. § 103 — U.S. Patent No. 10,037,467 B2

"Information processing system" · NEC Corporation · Inventor Takatoshi Kitano
App. 15/023,807 · PCT/JP2014/003631 · Priority JP 2013-200065 (2013-09-26) · PCT filed 2014-07-09 · Issued 2018-07-31

Scope of this section: This is a technical/patent-analyst opinion on whether the claims are vulnerable to a § 103 obviousness challenge built from the prior art of record cited on the face of the '467 patent (the "Citations (20)" list and PTL 1). It is not a legal conclusion.

---

0. Data provenance and a flagged gap (read first)

The authoritative patent text I was supplied truncates the "Citations (20)" list mid-entry at US6263088B1 ("1997-06-19 2…"). My copy therefore names only six of the twenty examiner-cited U.S. references plus PTL 1:

⚠️ Gap: the remaining ~14 of the 20 cited references are not in my copy and are not analyzed here. Any of them could be more material than the six I can see. Verify the complete list at patents.google.com/patent/US10037467/en.

Do not confuse the "before-the-fold" art with the "after" art. The page's "Families Citing this family (10)" and "Cited By (1)" lists (Samsung US12003881B2, NEC's own US11250243B2, Eizo WO2015064671/672, Hikvision CN107679529B, Hitachi JP7039409B2, Grazper EP4099212A1, etc.) are not prior art to a 2013-09-26 priority date — the earliest of them (Eizo) has priority 2013-11-01, five weeks after the '467 priority. They are § 102(a)(2)-irrelevant and cannot support a § 103 ground here. Likewise, the DELL IPRs and 2014–2015 CCE v. NEC IPRs discussed in the PTAB section of this record involve different patents and supply no art for the '467.

Cross-reference note (no contradiction): the earlier "Patent summary" section flagged an uncertainty whether issued claim 1 expressly recites a "store in a storage unit" step. That ambiguity is immaterial to this § 103 analysis — the association step (d) is present in claim 1 regardless, and the "store" language is express in claims 13 and 15, which are analyzed below on their own terms.

---

1. Governing framework and the person of ordinary skill

Effective filing date / critical date: 2013-09-26 (JP priority). All references above published 1992–2012 and are squarely § 102(a)(1) prior art.

POSITA (proposed): a person with a bachelor's degree in electrical engineering, computer science, or equivalent, and 2–3 years of experience in computer vision / video-content analysis, including object detection and tracking, facial-feature extraction, and feature-based image retrieval. This is the profile the § 101 opinion in NEC v. Anker effectively assumed when it credited the "detection, extraction, selection, and association process" as a technical image-processing workflow (see Dkt. 122 discussion in this record).

The single dispositive question. Claim 1 is not a claim to detecting objects, nor to extracting features, nor even to "best-shot" selection in the abstract — each of those was ancient by 2013. Its only plausibly distinguishing limitation is the per-element selection of the frame: a frame-selection criterion "set in advance for each of the object elements" (the specification's "modal" — face, clothes, sex/age), with the feature quantity of that element tied to that frame's identifying information. Everything else in claim 1 is old. The obviousness case therefore rises or falls on whether the art renders obvious applying best-frame selection on a per-object-element basis.

---

2. Claim 1 — element-by-element mapping to the art of record

On this mapping, limitations (a), (a′), (b) and (d) are each met by references cited on the face of the patent itself — i.e., the examiner already considered this art. If the challenged claims are to fall, the winning argument is the combination that supplies (c).

---

3. The centerpiece: why (c) — per-element frame selection — is obvious

3.1 What PTL 1 already teaches

As the '467 specification itself summarizes (Background Art), PTL 1 (JP2012-123460) discloses:

• detecting a specific event in a moving image;
• selecting, among the frames of the detected event, a frame "having a highest value regarding a predetermined index" as a best-shot image — its own examples being largest face area or face oriented frontally;
• extracting the face feature quantity from that best-shot frame; and
• searching the moving image using that feature quantity.

That is claim 1(a)-(b) and (d) plus a frame-selection-criterion-driven best-frame extractor for one element. The only thing PTL 1 lacks is doing it for more than one element.

3.2 What US5864630A already teaches

US5864630 ("Multi-modal method for locating objects in images," AT&T) is the natural completing reference. It expressly:

• receives a sequence of video frames (patents.google.com/patent/US5864630A/en);
• locates objects by running multiple modality channels (shape, color, motion) in parallel, each producing an intermediate feature representation of the same tracked features — head outlines and facial features;
• invokes/deactivates channels and reassesses results across frames to balance speed and accuracy; and
• fuses channel outputs into "a single list of likely head positions and the locations of facial features."

So US5864630 supplies precisely the missing ingredient the '467 specification calls a "modal": a system built around a plurality of independently-detected object elements, each with its own detection method and its own extracted feature representation.

3.3 The combination and the motivation to combine

Ground 1 (primary): PTL 1 + US5864630.

A POSITA reading PTL 1 would recognize that its best-shot selection is a quality-control step for the face, and would recognize that the identical quality problem recurs for any other element it might want to index. US5864630 provides the architecture in which several such elements already exist as first-class citizens, each with its own channel. The motivation is supplied by the references and by common sense (KSR):

1. Same problem, same solution, different element. PTL 1 selects a frame that maximizes face quality. The reason is that feature extraction is degraded by blur, area, and pose. Those same degradations (occlusion, orientation, motion blur) apply to clothes/color and to any other element; a POSITA would apply PTL 1's own best-frame criterion to those elements "to obtain a better feature." KSR: "if a technique has been used to improve one device, and a person of ordinary skill… would recognize that it would improve similar devices in the same way, using the technique is obvious."
2. US5864630 makes per-element treatment the design premise, not an afterthought. Because AT&T's channels already produce separate feature representations for head vs. facial features and already calibrate/reassess per element, selecting a best frame per element is a predictable, mechanical extension. A channel that is "often sufficiently reliable to run by itself for several frames" already contemplates element-specific frame handling.
3. Express design incentive already in PTL 1's own field. PTL 1's chosen indices (area, frontal orientation, blur) are frame-quality metrics; once quality is the objective for one element, the objective is the same for the rest.
4. The patent's own "problem" concedes the gap is one of application, not insight. The specification states PTL 1's shortfall is that it selects a best shot "solely regarding a face, and it is unclear whether a best shot image is determined for a search object other than a face." That is a complaint about incomplete application of a disclosed technique to other disclosed elements — the classic § 103 fact pattern, not a leap of invention. Cf. KSR (predictable use of prior-art elements).

Ground 1 mapping onto claim 1: US5864630 supplies (a), (a′) (multiple object elements: head + facial features) and (b) (per-channel feature representations) and the frame handling; PTL 1 supplies (c) (best frame by predetermined index) and (d) (extract feature from the best frame for subsequent search). Combined, all limitations are met.

3.4 Alternative and supplementary combinations (same claim element)

Ground 2: PTL 1 + US5689575A (Hitachi). US5689575 detects and tracks, per frame, a facial area and discrete facial parts — mouth area, eye area, and fingertip — determining "states for them according to their locations, orientation, open/shut status," i.e., per-element attributes at the frame level (patents.google.com/patent/US5689575). Combined with PTL 1's best-frame-by-index technique, a POSITA would select the frame that best shows each facial part (the frame where the mouth is most open/unoccluded, etc.) — again a per-element frame criterion, motivated by the same quality rationale. This is a particularly clean secondary reference because it also supplies the position information later claimed in claim 9.

Ground 3: PTL 1 + US6185314B1 and/or US6263088B1 (NCR). US6185314 evaluates region clusters per video frame to decide whether they are objects (persons) to be tracked, generating per-cluster features and confidence values; US6263088 tracks object movement and maintains identity across frames (patents.google.com/patent/US6185314). These provide (a)/(a′)/(b)/(d) and, importantly, the frame↔feature indexing and object identity (track) association that claims 1, 8 and 9 need. Combining with PTL 1 yields the per-element best-frame storage.

Ground 4 (detection backbone): US5103305A (Toshiba). For the bare "detect and track a moving object in image frames sequentially obtained" limitation, US5103305 discloses frame-by-frame moving-object detection in an image sequence and features extracted per frame; it is the least load-bearing reference and would be cited only to shore up (a).

3.5 Why the "the prior art selects a best frame only for a face" defense fails

NEC's expected rebuttal is its own Background-Art framing: no reference selects a best frame "for a search object other than a face." Three answers:

• Missing a single element ≠ non-obvious when the element is a predictable application of a disclosed technique. KSR, and the "familiar elements according to known methods" line of cases.
• US5864630 already treats element-specific detection/feature extraction as routine. Merging PTL 1's one-element best-frame selection into a multi-element framework is combination of references that address the same field and the same problem. KSR; In re Kahn.
• The specification's own advantages are the motivation. The '467 spec touts two benefits of per-modal best-frame selection: higher feature quality and reduced storage / faster search (only best-frame features are stored). Those are ordinary efficiency/quality improvements that a POSITA would have expected, satisfying the "reasonable expectation of success / predictable result" requirement rather than overcoming it.

---

4. Dependent claims 1→2–10

Net: every dependent claim adds either (i) a routine design choice (frame interval, temporal window, per-range count) or (ii) a known skill from the same references (track identity, part-position storage, feature-based search). None appears to carry an independent inventive step.

---

5. The search-side and method claims (11, 12, 13–14, 15–16)

• Claim 11 (search a store holding per-element features from criterion-selected frames) is PTL 1's search step generalized to a pre-populated store, plus the NCR feature/indexing infrastructure (US6185314, US6263088). It claims no capture machinery, so it is even more exposed: the "unconventional" aspects the § 101 opinion credited live on the indexing side (claims 1/13/15), and here they are simply presupposed as the storage contents.
• Claim 12 mirrors claim 7 (different-frame fallback) — same art and same rationale.
• Claim 13 (CRM/program) and Claim 15 (method) recite the same four functional units as claim 1; the § 103 analysis at § 2–3 applies verbatim. Claim 15's "before and after extracting the feature quantity" language merely states that selection may precede or follow extraction — a timing/order option the '467 specification itself presents as interchangeable ("the order of the process is not specifically limited"), i.e., an admitted design choice.
• Claim 14 (program further implements the search unit) → same as claim 6.
• Claim 16 (method further comprises searching) → same as claim 6.

---

6. Rebuttal landscape and residual strength of the claims

NEC's strongest non-obviousness themes (and my assessment):

1. "No reference selects a best frame for an element other than a face." — Factually accurate as to each individual reference, but legally weak under KSR: it is a predictable application of PTL 1's disclosed technique to elements that US5864630 (and US5689575) already show are separately detected and separately featured. Expect NEC to argue "teaching away" or "no reason to modify"; the counter is that PTL 1 and the multi-element references share the same field, same problem, and same quality rationale, and US5864630's channel architecture makes the modification a mere configuration change.
2. Secondary considerations. None is evident in the record I hold; the asserted claims appear to be practiced in NEC's own products and licensed to no one, so a nexus-based commercial-success story would be hard to build. No unexpected results are described beyond ordinary quality/efficiency gains.
3. § 101 survives. The June-2026 denial of Anker's JOP motion (Dkt. 122) held claim 1 is not directed to an abstract idea and "lacks a true human analog." That is an eligibility ruling only and is orthogonal to § 103 — indeed, the court's description of the claims as a concrete, multi-step image-processing workflow can just as easily be read as describing a predictable combination of known image-processing steps, which favors obviousness. Do not over-read the § 101 win as a validity shield.

Honest statement of residual strength. The six visible references each sit at 1992–2001 and none individually discloses per-modal best-frame selection; PTL 1 (2012) supplies the best-shot concept for one modal. A § 103 case therefore requires a genuine two- or three-reference combination (PTL 1 as the primary reference, US5864630/US5689575 as the multi-element enablers, NCR as the store/index/identity layer). Such a combination is legally viable and, in my view, more likely than not to yield a prima facie case against claims 1, 6, 7, 8, 9, 11 and their dependent counterparts — but it is not a lay-up, and the strength depends materially on the ~14 cited references I could not see, any of which might either reinforce (e.g., a second best-shot reference) or weaken (e.g., art already showing per-modal best-frame selection would strengthen the case) the analysis.

---

7. Confidence and caveats

• High confidence in the characterization of each named reference (verified against Google Patents / EPO family pages / FreePatentsOnline this session): US5864630 = multi-modal, multi-feature locating with per-channel feature representations; US6185314/US6263088 = NCR object tracking with per-region-cluster features and track identity; US5689575 = Hitachi facial-area + facial-part detection with per-part position/area; US5103305 = Toshiba frame-by-frame moving-object detection.
• Moderate confidence in the combination strength. The motivation-to-combine rationale is well-grounded, but I have not retrieved the actual § 102/§ 103 rejection record from the file wrapper (IFW/PTAB), and I did not independently verify PTL 1's full text — its content is taken from the '467 specification's own Background-Art characterization, which is authoritative for this analysis but is the patentee's characterization.
• Gap restated: the "Citations (20)" list in my authoritative source is truncated after US6263088B1; ~14 cited references are unexamined. Likewise, I did not import the post-2013 "Families Citing this family" or "Cited By" documents — they are not prior art to a 2013-09-26 priority and would be an error to cite.
• No fabrication: every reference identifier above is quoted exactly as it appears in the record; where a date or assignee is stated, it is from the cited source. Any reference not shown, I did not find.

Bottom line: Under § 103, the '467 claims are vulnerable to at least four workable combinations, all anchored on PTL 1 (JP2012-123460) as the primary reference teaching best-shot/best-frame selection by a predetermined index, combined with a multi-modal detector/feature extractor (US5864630A, alternatively US5689575A) that supplies the "plurality of object elements," and with US6185314B1 / US6263088B1 supplying tracking, per-element feature storage, frame↔feature indexing, and object-identity association. The motivation is a predictable quality/efficiency improvement that the '467 specification itself concedes PTL 1 fails only because it was applied to one element — which is the hallmark of an obvious, not an inventive, extension.