US 5835667

I'll research this patent number across the sources you specified.

Let me run a few more targeted searches to check for litigation and confirm the bibliographic record.

Research Findings — US Patent No. 5,835,667

Sources consulted and results

• Google Patents (primary full-text record, fetched 2026-09-27): https://patents.google.com/patent/[US5835667](/patent/US5835667)/en
• Espacenet bibliographic record: https://worldwide.espacenet.com/publicationDetails/biblio?CC=US&NR=[5835667A](/patent/5835667A)&KC=A&FT=D
• CAFC 2026 dockets: My searches of Federal Circuit 2026 materials returned no docket, appeal, or opinion referencing 5,835,667. The only 2026 CAFC-related results were unrelated matters (e.g., Taction v. Apple haptics, Ascendis v. BioMarin), with no connection to this patent. I could not identify any litigation or appeal involving this patent, but I cannot rule out untracked district-court or administrative proceedings — I have no authoritative negative-search confirmation.

Bibliographic summary (per Google Patents / Espacenet)

Identifier-consistency note (no auto-correction applied): Several third-party documents in the search results cite this patent with a different first-listed inventor — e.g., EPO search reports render it "US 5 835 667 A (MAULDIN MICHAEL L ET AL)," while others render it "US 5835667 A (WACTLAR HOWARD D [US] ET AL)." The authoritative front-page data (Google Patents and Espacenet) both list Wactlar first. I flag this as a citation inconsistency in the citing documents rather than silently reconciling it.

Abstract (verbatim)

"An apparatus and method of creating a digital library from audio data and video images. The method includes the steps of transcribing the audio data and marking the transcribed audio data with a first set of time-stamps and indexing the transcribed audio data. The method also includes the steps of digitizing the video data and marking the digitized video data with a second set of time-stamps related to the first set of time-stamps and segmenting the digitized video data into paragraphs according to a set of rules. The steps of storing the indexed audio data and the digitized video data with their respective sets of time-stamps is also provided. The method also includes the step of passing the transcribed audio data through a natural language interpreter before indexing the transcribed audio data. A method and apparatus for searching the digital library is disclosed."

Independent claims — plain-language overview

Claim 1 — Method of creating an independent digital library (independent, method)
Steps: (a) transcribe the existing audio and time-stamp it with a first set of time-stamps; (b) index the transcribed audio; (c) digitize the video and time-stamp it with a second set of time-stamps related to the first set; (d) segment the digitized video into "video paragraphs" according to a set of rules based on scene characterization of the video images and on processing of the audio data; and (e) store the indexed audio plus segmented video with their time-stamps, yielding a library "which can be accessed through said indexed audio data without returning to the existing audio data and video images." In plain terms: build a self-contained, searchable multimedia database from source tape by aligning speech transcripts and video on a common time axis, then cutting the video at content-based (not merely time-based) boundaries, so later searching works off the stored index rather than the original source media.

Claim 4 — Apparatus (independent, apparatus)
The counterpart of claim 1 in means-plus-function form: means for transcribing and time-stamping audio; means for indexing it; means for digitizing and time-stamping video relative to the audio stamps; means for storing a rules set based on scene characterization and audio processing; means for segmenting the video into video paragraphs per those rules; and means for storing the indexed audio and segmented video with their time-stamps to create the independently-accessible library.

Claims 2–3, 5–6 (dependent): add the natural-language interpreter step/means before indexing, and the feature that the interpreter updates the set of rules used for segmentation.

Additional dependent claims visible in the record: 7–9 (icon generation; "intelligent moving icons"/imicons; data-dependent heuristics), 10 (video compression before storage), 11–13 (the speech-recognition pipeline: forward time-synchronous pass producing candidate word occurrences, reverse pass producing alternate begin times, N-best hypothesis generation via approximate A*, and selecting the best-scoring hypothesis; trigram language model).

Uncertainty / coverage caveat: The claim text available to me is truncated mid-claim 13 ("…produced using a reve…"). The patent carries 34 claims total, so claims 14–34 — which, per the specification's statement that "a method and apparatus for searching the digital library is disclosed," likely include one or more additional independent search/retrieval claims — are not in my authoritative source text. I therefore cannot confirm the full set of independent claims beyond claims 1 and 4. I will not guess at their content.

Contextual notes (from the specification)

• The patent is the foundational Informedia digital-video-library patent; the specification names CMU's Sphinx-II continuous speech recognizer (Huang 1993) and the Scout full-text retrieval system.
• It expressly incorporates by reference a concurrently filed companion application, Mauldin et al., "System and Method for Skimming Digital Audio/Video Data" — which corresponds in the search results to US 5,664,227 (Mauldin et al., issued Sept. 2, 1997), also assigned to Carnegie Mellon.

I'll search for litigation involving US patent 5835667 across patent litigation databases and court sources.

Initial searches show only patent citations and Office actions, no litigation. Let me check litigation-specific databases more directly.

Let me search directly in litigation databases and PACER/CourtListener for this specific patent.

Known Litigation Involving U.S. Patent No. 5,835,667

Summary: No litigation (district court, ITC, or PTAB) involving U.S. Patent No. 5,835,667 was found.

I searched patent-litigation and related sources (Google Patents legal-status/prosecution record, Unified Patents materials and the litigation references surfaced therein, PACER/CourtListener-indexed results, EP search reports and family data, and general web queries on the number in both plain and comma-formatted styles). Every "hit" for the number was a prior-art citation, an international search report (ISA/EPO) reference, or a family-listing entry — not an assertion of the patent.

What the record actually shows for US 5,835,667

Because the patent expired on/around November 10, 2015, it could not have been asserted in any suit filed after that date (absent back-damages for pre-expiration conduct, which still would have required a complaint within the limitations period). No evidence of such a suit was located.

Important caution — do not conflate similar numbers

The strict "interpret identifiers literally" rule matters here, because my searches surfaced several different patents whose numbers are easily confused with 5,835,667. These are NOT US 5,835,667 and must not be reported as litigation against it:

• U.S. Patent No. 5,677,566 (the "'566 patent") — a DRAM/semiconductor patent asserted by Nanya Technology against Elpida and Kingston in an ITC Section 337 complaint (filed Nov. 21, 2011). This is Nanya's '566 patent, not CMU's '567 patent.
• A "566 patent" in W.L. Gore & Associates v. Garlock (Fed. Cir.) — a PTFE/plumbing-tape patent, again unrelated.
• Trademark Registration No. 5,835,667 (BRYHALI, Bausch Health) — a trademark registration, not a patent.
• Marvell v. Carnegie Mellon University litigation and the CMU v. LSI/Avago and Marvell Semiconductor suits involve CMU's disk-drive read-channel patents (e.g., the '839/'889/"180" patents) — not US 5,835,667.

Confidence and caveats

• High confidence that no publicly indexed district-court, ITC, or PTAB proceeding names US 5,835,667 as an asserted patent. The patent appears in the literature only as prior art (it is cited as an "X" reference in later EP/PCT search reports, e.g., EP 3 329 386 and WO 2000/072186, and in EP 1 278 134 as a category-A reference).
• Caveat: I could not complete exhaustive direct queries of the Unified Patents litigation case list, Docket Alarm, or full PACER for the exact number because I hit the tool-step limit. If you require a court-record-certified negative, the definitive steps would be: (1) Unified Patents Litigation portal search on 5835667; (2) Docket Alarm / PACER "patent number" search filtered to 5835667; (3) Lex Machina / Westlaw Litigation Analytics on the patent number; and (4) the USPTO PatentCenter "Litigation" / assignment tabs for US 08/324,076. On the public-record evidence gathered, however, the answer remains: no known litigation.

If your interest is in CMU's video-library technology specifically, note the patent was never (per available records) commercialized through an assertion campaign, and it lapsed/expired in 2015 — which is consistent with the absence of litigation.

I'll verify the canonical "no proceedings" default against public sources before writing up.

No PTAB hits surfaced. Let me run additional targeted checks — including litigation-driven IPRs and secondary-source databases.

Proceedings overview

Total AIA trial proceedings on US 5,835,667: zero. The canonical structured source ("PTAB proceedings on file") reports no AIA trial proceedings for this patent as of the most recent USPTO Open Data Portal ingest, and my independent web verification surfaced no IPR, PGR, or CBM petition, institution decision, Final Written Decision, or PTAB appeal referencing this patent. Defensive posture for a defendant: there is nothing to cite and nothing to rely on — but that is not because the patent is a hardened survivor of PTAB attack. It is because no one ever needed to file. The patent is expired (anticipated expiration 2015-11-10) and appears to have been licensed rather than litigated, so it never attracted the assertion campaign that drives IPRs. Any defense strategy must be built on expiration and the § 286 damages bar, not on PTAB outcomes.

No proceedings to report

I cannot populate the per-proceeding template (number, petitioner, panel, grounds, FWD, appeal) because there are no proceedings, and I will not invent numbers. Search coverage and what each check returned:

Two categorical observations that explain the empty docket (both grounded in the record, not speculation):

1. PGR is unavailable as a matter of law. Post-grant review applies only to patents whose claims have an effective filing date on or after 2013-03-16. US 5,835,667 has a filing/priority date of 1994-10-14 — two decades too early.
2. CBM is unavailable as a matter of law and time. Covered Business Method review was limited to pre-AIA-effective-filing-date patents and sunset on 2020-09-16. That door is closed regardless of whether this patent would otherwise have qualified.

That leaves IPR as the only theoretical AIA vehicle left, and it is available even against an expired patent — but with no injunction exposure and an expired claim term, the economic case for an IPR is essentially nil.

Strategic summary

Claim status: CANCELED — none. SUSTAINED — none (never challenged). UNTESTED — all 34 claims (claims 1–4 independent; claims 2–3, 5–6, 7–13 and the balance dependent, per the record available to me). Critically, no AIA proceeding has narrowed this patent by a single claim. Because no FWD exists, there is no claim-level invalidation record, no judicially or administratively adopted claim construction, and no estoppel to work with or against. A defendant gets no free kill-shot, but also inherits no adverse PTAB precedent.

Estoppel landscape: not applicable. The § 315(e)(2) estoppel bar is triggered by a petitioner's participation in an instituted IPR that reaches a final written decision. With no proceedings, no party is estopped from anything, and all prior-art grounds remain theoretically available — in an IPR, in a district court § 282 defense, or in an ex parte reexamination. Practically, the more useful limitation is temporal, not estoppel-based: any invalidity or damages challenge is muted by the patent's expiration on 2015-11-10, which predates the six-year damages lookback from a 2026 filing by roughly a decade.

Pattern signals: no pattern, because no activity. There is no repeat petitioner, no aggressive Patent Owner appeal history, and no evidence of a defensive aggregator (Unified Patents, RPX, etc.) in the chain. The licensing posture appears instead to have been commercial and consensual: a public "Amended and Restated License Agreement between Carnegie Mellon University and Sonic Foundry Systems Group, Inc." covering software and patents "developed under the Informedia Project" (https://contracts.justia.com/companies/sonic-foundry-inc-1659/contract/[1122895](/patent/1122895)/). I flag that I did not independently confirm that US 5,835,667 is enumerated in that agreement — it appears to relate to the same Informedia technology, and the agreement's existence is consistent with the absence of PTAB disputes, but I will not overstate the link.

Contradiction / caveat to carry forward from the earlier section: the claim text available to me is truncated mid-claim 13, and claims 14–34 are not in the authoritative source text. That matters here for a specific reason: if a demand letter cites claims 14–34, I have no basis to assess their scope, and no PTAB record to tell me whether anyone ever construed or invalidated them. Do not assume the dependent claims are narrower in any way that helps.

Recommended next steps

If you are a defendant and no PTAB activity exists — say so plainly, and pivot to the real defenses. The absence of IPRs is itself a signal, but not the signal it usually is. Ordinarily, "no IPRs" suggests a patent too weak or too unimportant to bother with. Here it reflects a 1994-priority, university-owned, already-expired patent that was licensed rather than enforced. Recommended order of attack:

1. Expiration + § 286 time bar (strongest ground). The patent expired 2015-11-10. Any recovery for infringement occurring more than six years before a complaint is filed is barred by 35 U.S.C. § 286. For a suit filed on or after 2019-11-10, every possible act of infringement predates the six-year window by construction — i.e., the entire damages period is arguably non-recoverable. Also confirm there is no residual § 271(e)(2)/Hatch-Waxman or § 287 wrinkle, and check whether any pre-2015 licensee or co-owner standing defect exists (Carnegie Mellon ownership/co-ownership standing has been litigated elsewhere and is a recurring university-patent vulnerability). (This is an argument to develop with counsel — § 286 is a limitation on recovery, and equitable doctrines can bear on it — not a certainty I can certify.)
2. Invalidity, if the damages picture is somehow live. With no estoppel and no prior PTAB record, your § 102/§ 103 defenses and any ex parte reexamination request are unrestricted. Note that the specification itself concedes the prior-art baseline heavily — it frames the invention largely as an integration of known techniques (Sphinx-II speech recognition, per Huang 1993, incorporated by reference; the Scout retrieval system; DCT-based key-frame detection; Kalman filtering; the "factorization" 3-D approach). That framing is fertile § 103 territory.
3. Check the companion patent. The specification incorporates by reference Mauldin et al., "System and Method for Skimming Digital Audio/Video Data" — which the prior section maps to US 5,664,227 (Carnegie Mellon). If an assertion implicates skimming functionality, confirm which patent is actually being asserted.
4. Verify the negative yourself, once, on the record. Use PTAB E2E (https://e2e.uspto.gov/) and the PTAB decisions search (https://developer.uspto.gov/ptab-web/#/search/decisions), plus Docket Alarm's PTAB docket (https://www.docketalarm.com/cases/PTAB/) and CourtListener (https://www.courtlistener.com/) for any CAFC appeal. I searched several of these indirectly and found nothing, but I could not execute a direct E2E docket pull, so my "zero" rests on the structured ODP block plus corroborating searches rather than a live E2E screenshot.

Net bottom line: there is no FWD to link to and no canceled claim to quote — I will not manufacture either. The patent's defensive weakness is temporal, not adjudicative: it expired on 2015-11-10 and no AIA tribunal has ever touched it.

I'll research the assignment record for this patent across the specified sources.

Let me search more specifically for the assignment records and any litigation involving this patent.

Let me run targeted searches on the assignment database, litigation, and the companion patent to confirm the chain.

Assignment / Ownership Reconstruction — US 5,835,667

Scope note and sourcing limitation (read first). I attempted to pull the recorded assignment pages from USPTO Assignment Center (assignmentcenter.uspto.gov) and its legacy mirror (assignment.uspto.gov), plus PEDS/ODP, and to cross-reference RPX/Unified court-and-defendant records. My web tooling did not return a live Assignment Center result page for this patent, and the attorney/agent correspondent-of-record field was not retrievable in this session. Accordingly, I do not have a verified reel/frame number to cite, and I will not fabricate one. What I can state with confidence comes from the Google Patents legal-events record for US5835667A (fetched 2026-09-27, https://patents.google.com/patent/US5835667/en), which mirrors the single recorded assignment to the original assignee and shows no post-issuance assignment, security-interest, merger, or name-change events. Treat the reel/frame redaction below as a coverage gap, not as an assertion that no document exists.

Inventors

All seven named inventors were, at the 1994-10-14 filing date, personnel of Carnegie Mellon University (CMU), Pittsburgh, PA — consistent with the patent's "Assignment of Interest" recording to CMU.

Pattern note (potential, not a fire-sale tell). This is a single-employer, single-assignee inventor group — not the "all inventors depart within 12 months" pre-fire-sale pattern the task asks about. The one departure-adjacent datum worth flagging: Michael L. Mauldin is widely credited with founding Lycos, Inc. in 1994 while associated with CMU, and is reported to have moved to Lycos full-time thereafter. I have not independently verified the exact departure date or whether it bears on title, so I flag it as an item to confirm rather than a finding; in any event an inventor's personal move does not disturb a corporate assignment already made to CMU. No other inventor is known to have left CMU in the relevant window.

Original assignee

Carnegie Mellon University (original and current assignee per Google Patents and Espacenet — both list CMU, no later assignee).

• Primary line of business: nonprofit research university (computer science, robotics, speech/language, software engineering).
• Product embodying the claims: CMU is a university, not a commercial vendor. The patented technology was embodied in the research Informedia Digital Video Library system (the specification names CMU's Sphinx-II recognizer and Scout retrieval system, and the work was funded under NSF Cooperative Agreement IRI-9411299). Informedia was a research/演示 system, not a shipped commercial product.
• Current status: operating — CMU remains an ongoing, solvent institution. No bankruptcy, dissolution, or acquisition of the assignee is implicated in this chain.
• Companion family patent: Mauldin et al., "System and Method for Skimming Digital Audio/Video Data" (US 5,664,227, issued 1997-09-02), also CMU — incorporated by reference in this patent's specification. It shares the CMU ownership profile described here.

Assignment timeline

Chronological record as reflected in the Google Patents legal-events feed for US5835667A. Only one recorded assignment event appears; the reel/frame and correspondent fields are not recoverable from my sources and are marked accordingly.

• Executed 1994-10-14 (application filing date; execution date not separately confirmed) / recorded 1995-03-22 — Reel NNNNNN/NNNN (not retrieved — see scope note)

  • Conveyance: Assignment (Google Patents event code "AS – Assignment"; description "ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS)")
  • Assignor: Takeo Kanade; D. Raj Reddy; Michael A. Smith; Michael G. Christel; Scott M. Stevens; Howard D. Wactlar; Michael L. Mauldin (individually, as the seven inventors)
  • Assignee: Carnegie Mellon University
  • Correspondent: not retrieved in this session — cannot confirm the attorney/agent of record or firm. No recurrence assessment is possible.
  • Context: intake assignment — each inventor assigned his pre-filing rights in the invention to his employer/university at the outset. This is a standard institutional employment/invention-assignment, i.e., the original assignment, not a transfer.

• No further events. Google Patents shows no post-issuance assignment, security agreement, merger, change of name, license, or release recorded against this patent. Anticipated expiration 2015-11-10, and legal status is Expired – Lifetime (term ran full length; no maintenance-fee lapse at issue).

Cross-reference result: I found no evidence in any source of an assignment to a licensing entity, NPE, defensive aggregator, or SPV, and no litigation, IPR/PGR, or Federal Circuit matter naming this patent. My prior-pass note that no CAFC 2026 docket references 5,835,667 stands; I still cannot supply an authoritative negative-search certification for all district-court/administrative proceedings, so I state the absence as "not found," not as "does not exist."

Timeline diagram

timeline
    title Ownership of US 5835667
    1994 : Invented at Carnegie Mellon
         : Application filed
    1995 : Assigned to Carnegie Mellon
    1998 : Patent issued
    2015 : Patent expires full term

NPE / troll-pattern signals

Assessed against the record above. Because there is only the original intake assignment and no post-issuance chain, most signals resolve to not present on the evidence available.

1. Shell-entity transfer — Not present. No assignment from CMU to any "IP / Holdings / Licensing / Ventures" entity appears in the record; the sole recorded assignee remains Carnegie Mellon University itself.

2. Known asserter in the chain — Not present. No current or prior assignee matches any public NPE list (Acacia, Marathon, IV, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation, Spangenberg entities, or any RPX/Unified high-frequency plaintiff). Assignee = CMU throughout.

3. Repeat correspondent across the chain — Unclear (data unavailable). A single-link chain makes "recurrence" moot, and the correspondent-of-record could not be retrieved this session. No flag either way.

4. Cascading transfers — Not present. No consecutive assignments; nothing moved in <24 months.

5. Pre-litigation transfer — Not present. No assignment precedes any suit, because no suit naming this patent was found and no transfer occurred after issuance.

6. Bankruptcy fire-sale — Not present. CMU did not file bankruptcy; the patent was not sold in any proceeding. (Contrast the Kodak/Nortel/Polaroid-style scenarios named in the task.)

7. Privateering — Not present. No operating company transferred the patent to an NPE for assertion; no SEC/Patent Progress/EFF coverage surfaced.

8. Defensive aggregator (anti-NPE) — Not present. The chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. It terminates at the original university assignee, which simply held the patent to expiration.

Verdict

Insufficient data — in the specific sense the rubric anticipates: only the original assignment is on record. This is a benign instance of that category, not a hidden chain.

Justification (2–3 sentences): The Google Patents legal-events record for US5835667A shows exactly one ownership event — the 1995-03-22 assignment of assignors Kanade, Reddy, Smith, Christel, Stevens, Wactlar, and Mauldin to Carnegie Mellon University — with no subsequent assignment, merger, security interest, or name change, and the patent expired full-term on 2015-11-10. Assignee therefore remained CMU from filing to expiration, so no NPE signal is present and none of the shell-entity, cascading-transfer, pre-litigation, privateering, or defensive-aggregator tests is met. The residual uncertainty is purely evidentiary: I could not retrieve the reel/frame or correspondent-of-record from Assignment Center in this session, so the single link cannot be cited to a recorded document number and should be verified directly at the USPTO Assignment Center search page — https://assignmentcenter.uspto.gov/ (search "5835667"; legacy mirror https://assignment.uspto.gov/patent/index.html).

Verification steps I recommend for a clean certification: (1) pull the assignment abstract for 08/324,076 / 5,835,667 at Assignment Center to capture reel/frame and correspondent; (2) confirm no security-interest or license records exist; (3) run the patent number against RPX and Unified Patents asserter directories and Docket Navigator/LexMachina for any assertion history.

I'll research the prior art references cited in US 5,835,667, prioritizing the authoritative patent record.

I need the actual "References Cited" / "Patent Citations" table from this patent. Let me query for it more specifically.

The fetched full text omitted the front-page citation list. Let me pull the citations from database mirrors (FreePatentsOnline, Justia) and the USPTO record.

Important research limitation — read first

I was unable to retrieve the authoritative front-page "References Cited / U.S. Patent Documents" table for US 5,835,667. The Google Patents full text supplied in this matter (and the mirror I re-fetched at https://patents.google.com/patent/US5835667/en) renders the specification, claims, and classification but omitted the patent-citation table. Espacenet's bibliographic view (https://worldwide.espacenet.com/publicationDetails/biblio?CC=US&NR=5835667A&KC=A&FT=D) likewise omits the examiner-cited list. My repeated queries for the citation table returned forward citations (later documents citing '667) and search-report family annexes — not the list of references cited in '667.

Under the operating rule to not fabricate, I will not enumerate specific patent numbers as "the references cited in 5835667" when I could not verify them. Anyone who needs that list verbatim should pull the front page (PDF) of US 5,835,667 from USPTO PatentCenter / patentimages, which is the only place the examiner's References Cited block is authoritative.

What I can do — and do below — is (a) confirm the record, (b) give the references the patent itself expressly cites/incorporates (verifiable from the text I have), (c) list the documents that cite '667 (verifiable from search reports I retrieved), and (d) explain why the latter cannot be §102 art against '667.

1. USPTO record for 5835667 (literal, confirmed)

Critical date for §102: Because '667 was filed 1994-10-14, this is a pre-AIA patent. Only references with an effective date before 1994-10-14 can be §102(a)/(b)/(e)/(g) art. Anything published after that date — including every search-report citation I found — cannot anticipate '667.

2. References expressly cited/incorporated within US 5,835,667

These are verifiable from the '667 specification text itself (and are the only "citations by the patent" I can ground). Note: none of these is a patent cited on the front page; three are non-patent literature and one is a co-pending U.S. application incorporated by reference.

Caution on APP-1: The prior summary correctly flags that US 5,664,227 is the skimming companion. Because '667 incorporates it by reference, it should not be treated as §102 prior art against '667 — it is treated as if part of '667's disclosure. I flag this to prevent a mis-citation.

3. Documents that cite US 5,835,667 (forward citations — verified from retrieved search reports)

These are the only patent documents I could positively tie to '667 by number. They are forward citations, i.e., later applications for which '667 was cited as prior art — the reverse of what the task asks, but they are the citations the record actually shows:

These cannot be §102 prior art against '667: every one post-dates 1994-10-14. They are useful only as evidence that '667 is a recognized seminal reference in the content-based video indexing / transcript-search art — which is precisely why later examiners used it as an X/Y reference against others.

4. §102 anticipation analysis — what can and cannot be said

What I can state with confidence:

• No verified reference, on the evidence retrieved, anticipates claims 1 and 4. I could not obtain the examiner-cited list, so I cannot rule any reference in or out; I will not guess.
• No document of record appears capable of §102 anticipation of claim 1 as a whole. Claim 1 requires the combination of (i) audio transcription + time-stamping, (ii) indexing, (iii) video digitization + related time-stamping, (iv) content-based segmentation "into video paragraphs according to a set of rules based upon scene characterization of said video images and the processing of said audio data," and (v) storage yielding a library "accessed through said indexed audio data without returning to the existing audio data and video images." Anticipation under §102 requires all elements in a single reference; the retrieved candidates (later EP search reports) do not even qualify temporally. A reference disclosing only time-based or frame-count segmentation would not anticipate, because element (iv) is expressly content-based and multimodal.
• Claim 11–13 pipeline: the Sphinx-II description (NPL-1, 1993) describes the forward/reverse time-synchronous passes, N-best generation, and best-hypothesis selection. If those specific features are in the printed Huang 1993 reference, they could be §102 art against claims 11–13 — but I have not verified the reference's internal content, so I do not assert this.

What I explicitly decline to assert:

• Any mapping of specific U.S. patent numbers from '667's front page to specific claims, because I could not obtain that front page.

5. How to close the gap (definitive steps)

To produce the authoritative per-reference, per-claim §102 table the task requires, retrieve:

1. The US 5,835,667 front page PDF — https://patentimages.storage.googleapis.com/.../US5835667.pdf or USPTO PatentCenter for application 08/324,076 → "References Cited" block (U.S. Patent Documents + Foreign Patent Documents + Other Publications).
2. The microfilm/IFW "References Cited" and examiner's Form 892 for application 08/324,076 (USPTO PatentCenter → Documents tab).
3. Espacenet "Cited Documents" tab (distinct from "Also published as"): https://worldwide.espacenet.com/publicationDetails/citedDocuments?CC=US&NR=5835667A&KC=A&FT=D.
4. Cross-check categories against the EP 0 786 114 B1 prosecution file (the European counterpart), which contains an EPO search report listing the same art with X/Y/A relevance codes.

Once that front-page list is in hand, the §102 mapping is mechanical: filter to references dated before 1994-10-14, then test each against claim 1's five-element combination and claim 4's means-plus-function counterpart, then against dependent claims 2–3, 5–10 (NL interpreter/rules update, icons/imicons, compression) and 11–13 (ASR pipeline).

Bottom line

• US 5,835,667 confirmed as CMU's Informedia patent, filed/priority 1994-10-14, granted 1998-11-10, 34 claims, expired.
• I could not verify the examiner-cited prior-art list and therefore will not enumerate specific cited patent numbers as §102 art.
• The only references I can ground from '667 itself are Huang 1993 (Sphinx-II), Stevens 1992 (ALT Project), and the incorporated Mauldin et al. skimming application — with the important caveat that the last is not independent §102 art against its own incorporating patent.
• The forward citations (EP 3 329 386, EP 1 566 753, EP 1 262 881, EP 1 278 134, US 8,156,114, etc.) post-date '667's critical date and therefore cannot anticipate it; they instead confirm '667's role as prior art to later filings.

Obviousness Analysis — U.S. Patent No. 5,835,667 under 35 U.S.C. § 103

Preliminary caveat on the source material (please read first)

The materials supplied to me do not contain an explicit "Prior Art" section for US 5,835,667. The two previously generated sections (Patent summary; Litigation summary) and the full patent text you provided contain: bibliographic data, the abstract, the specification, a truncated claim set, and "Prior art keywords" (video; audio data; data; rules; time) — but no list of "References Cited," no examiner-cited art, and no EP/ISA search-report art list for this patent.

Because I cannot invent a document list that was not provided, I have built the analysis below on:

1. References whose content and dates I could verify through the searches run in this session (principally the reference list of the Informedia paper "Text, Speech, and Vision for Video Segmentation: The Informedia Project" (1995 output listing 10 references), and search-report entries surfaced for US 5,835,667).
2. Admissions in the patent's own specification, which are usable as prior-art evidence (they establish what was known to the art).
3. Well-known contemporaneous references in the same field.

Any specific reference I could not verify is flagged as such. Where I name a document but could not read its disclosure, I say so rather than assert what it teaches. If you have the actual "Prior Art" tab contents, supply them and I will re-run the mapping against those exact documents.

Also note: the patent has a 1994-10-14 priority date, so pre-AIA § 102/§ 103 governs. The critical dates are (a) the invention date (unknown to me) for § 102(a)/(g) art, and (b) 1993-10-14 for § 102(b) statutory-bar art. I flag date status for each reference.

1. Legal framework applied

Under Graham v. John Deere Co., 383 U.S. 1 (1966) and KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007), obviousness turns on: (1) scope and content of the prior art; (2) the level of ordinary skill; (3) the differences between the prior art and the claims; and (4) whether those differences would have been obvious to a PHOSITA, considering (a) predictable results, (b) design incentives/market forces, (c) interrelated teachings, and (d) whether the combination was "obvious to try."

Level of ordinary skill (proposed): A PHOSITA at the 1993–1994 timeframe would be a computer-science/electrical-engineering practitioner (or a small team) with a bachelor's degree plus ~2–5 years' experience in multimedia systems, and working familiarity with: continuous-speech recognition (at least the HMM/N-gram pipeline), information retrieval/inverted indexes, and video processing (scene-change detection, DCT coefficient analysis). The patent itself treats these as separate known arts and combines them — so the PHOSITA is a multimedia integrator, not a specialist in any one discipline. This is important because the claims are essentially integrative, which strongly favors obviousness under KSR's "familiar elements according to known methods" rationale.

Claim construction note: "Independent digital library," "video paragraph," "imicon," and "rules based upon scene characterization … and the processing of said audio data" are not terms of art. Under the broadest reasonable interpretation (and the specification's own usage), "video paragraph" = a content-bounded video segment, and the "rules" are the segmentation heuristics. Nothing in the claim requires any particular algorithm, so the claim scope is broad.

2. The prior art

2.1 References verified in this session

2.2 Reference I could name but not read (flagged)

2.3 Admissions in the patent's own specification (usable as art evidence)

The specification concedes the following were known, which narrows the "difference" inquiry dramatically:

• Transcription: Sphinx-II "known techniques in automated speech recognition" (and human transcription or closed captions offered as alternatives).
• Time-stamping: "any known technique for applying a time stamp."
• Indexing: "techniques generally known in the art … an inverted index."
• Video digitization: "techniques known by those skilled in the art."
• Key-frame detection by DCT coefficient changes: "as is known" / content-independent methods "detect big 'image changes.'"
• Image statistics/histograms for comparison: "proven to be useful."
• Natural-language interpreters: "based on known techniques … statistical techniques or expert systems"; Scout "and other natural language interpreters … are known."
• Retrieval models: "Boolean and vector-space models of information retrieval have been applied."
• Video-on-demand / constant-rate multimedia playback: discussed as the existing paradigm.

3. Claim 1 — element-by-element obviousness mapping

Claim 1 requires: (a) transcribe audio + first time-stamps; (b) index the transcript; (c) digitize video + second time-stamps related to the first; (d) segment video into video paragraphs per rules based on scene characterization AND audio processing; (e) store indexed audio + segmented video with time-stamps to create a library usable without returning to the source.

Claim 1 conclusion (proposed): Obvious over A1 + C1/C2 + B1 + D1/E1 (with F1, if its disclosure matches its title, as an even stronger primary reference). Every element was independently known; the asserted invention is the arrangement of them.

4. Motivation to combine (KSR factors)

1. Same field of endeavor. A1, B1, C1/C2, D1/E1, and F1 are all in multimedia/information-retrieval or audio-video processing. No field-crossing problem.
2. Predictable results. Combining (i) a time-aligned transcript index with (ii) a content-based video partitioner to yield (iii) a searchable stored library is a combination of known elements with no change in their respective functions. KSR, 550 U.S. at 416 ("familiar elements … according to known methods is likely to be obvious").
3. Market/design incentives — expressly recited in the patent's own background. The specification frames the need as: "New technology is needed to create, organize, and search the vast data libraries," the "Information Superhighway," and the inadequacy of video-on-demand's "store and replay." That is a documented design incentive to combine ASR + IR + video segmentation.
4. Interrelated teachings. The patent itself states that "only by the collaborative interaction of image, speech, and natural language understanding technology can the present invention … automatically populate, segment, index, and search." That is, the inventors themselves assert the arts were known to be combinable for this purpose.
5. "Obvious to try." Given a finite, identified set of known segmentation cues (scene change, camera motion, speaker change, audio/music change), trying the union of them to improve segmentation reliability (the very point the patent makes about FIG. 6) is routine experimentation.

5. Independent claim 4 (apparatus)

Claim 4 is the means-plus-function mirror of claim 1 (35 U.S.C. § 112 ¶ 6). It adds no substantive limitations beyond claim 1 and is obvious for the same reasons. Under § 112 ¶ 6 the corresponding structures are the general-purpose processors/software described in the specification (e.g., a "150 MIPS DEC Alpha workstation"), all conventional. A1 + B1 + C1/C2 + D1/E1 renders claim 4 obvious.

6. Dependent claims

Claims 2–3 and 5–6 — natural-language interpreter; interpreter updates the rules

• Differences: add (i) passing the transcript through an NLP interpreter before indexing and (ii) feeding segmentation rules from that interpreter to the paragrapher.
• Where taught/suggested: C2 (Mauldin text skimming/NLP retrieval), C1 (IR models), and the patent's admission that NLP processors "are known in the art." Using NLP output to derive content keywords (e.g., "football"/"scoreboard") as segmentation cues is the natural extension of B1's content-based segmentation plus C2's NLP.
• Motivation: using linguistic content to improve visual segmentation is exactly the "interrelated teachings" point; predictable result.
• Conclusion: Obvious over A1 + C1/C2 + B1 (and dependent claims 3/6 obvious as the automated feedback of the same NLP output that the reference already produces).

Claims 7–9 — icon generation; "intelligent moving icons"; data-dependent heuristics

• Differences: generate icons after segmentation; in particular "imicons" built with data-dependent heuristics.
• Where taught/suggested: B1 (key-frame/representative-frame selection) and E1 (representative frames for segmented video). The specification itself admits "representative frame" abstraction is "as is known" and that icons are "a combination of text and video." The imicon heuristic (select optimal frames/frame-rate/scene count) is optimization of a known representation.
• Caution: The terms "micon"/"imicon" are the applicants' coinages, and the skimming functionality is cross-referenced to the companion Mauldin et al. application (the search results indicate this corresponds to US 5,664,227, issued 1997-09-02). Because that application shares the same 1994-10-14 filing date, it is not § 102 prior art against this patent; the imicon-specific subject matter is therefore best attacked with B1 + E1, not with the companion case.
• Conclusion: Obvious over B1 + E1; claims 8–9 obvious as routine optimization ("obvious to try") under KSR.

Claim 10 — compress the digitized video before storing

• Where taught/suggested: the specification admits commercial formats were available and expressly names Intel DVI, MPEG, MPEG-II; E1 discloses compressed/digitized stored video.
• Conclusion: Obvious — compressing video for storage is the paradigmatic predictable design choice; no unexpected result.

Claims 11–13 — the ASR pipeline (forward time-synchronous pass → candidate word occurrences with start/possible end times; reverse pass → alternate begin times; N-best hypotheses via approximate A*; select best-scoring hypothesis; trigram language model)

• Where taught/suggested: A1 (Sphinx-II) and A2 (Hwang et al.: phone-dependent codebooks + adaptive language models). The specification itself describes this pipeline as "four known steps" and cites Huang (1993).
• Conclusion: These claims recite the admitted, published Sphinx-II pipeline almost verbatim. A1 alone (and A1 in view of A2) renders claims 11–13 obvious. Because A1 is incorporated by reference into the patent, it is unquestionably part of the art.
• Caveat: Claim 13 is truncated in my source ("…produced using a reve…"). I cannot opine on its precise full scope; the visible structure (a "reverse" pass) matches A1.

Claims 14–34 — NOT ANALYZED (insufficient text)

The full text I have truncates at claim 13; the patent has 34 claims. Per the specification's statement that "a method and apparatus for searching the digital library is disclosed," claims 14–34 likely include one or more independent search/retrieval claims (natural-language query → phonetic-space matching, recall/precision, ordering of "hits"). I will not guess at their content. If those search claims recite the phonetic partial-match technique described in the specification (the "self-fulfilling prophecies" / "self-fulfilling profit seize" example), the likely § 103 combination is C1 (Salton & McGill — Boolean/vector retrieval) + C2 (Mauldin — partial-match/string-similarity retrieval) + A1 (phonetic output), on the rationale that matching in phonetic space is a predictable extension of known partial-match/string-similarity retrieval when the corpus is an error-laden ASR transcript.

7. Secondary considerations / potential rebuttal

I found no evidence of record here that rebuts the prima facie case, but the following would be the applicants' likely arguments and my assessment:

• Unexpected results: The applicants would likely point to the integration's performance despite error-laden transcripts. However, the specification frames the phonetic-matching benefit as the expected consequence of combining known ASR output with known partial-match retrieval — not a surprising result. Weak.
• Commercial success / recognition: The Informedia project is well documented (search results show numerous CMU/NIST/TRECVID papers and a commercial spin-off, MediaSite). But there is no evidence in my record of a nexus between any such success and the specific claim elements, and the patent expired 2015-11-10 and was never asserted (per the previously generated Litigation summary). Weak-to-neutral.
• Teaching away: I found none. The reference B1 affirmatively teaches toward content-based segmentation, and the patent's own FIG. 6 teaches that combining cues improves reliability — the opposite of a teaching away.
• § 103(c) / common ownership: Some references (e.g., C2, Mauldin 1989) and the companion Mauldin et al. application share CMU authorship/ownership. Under pre-AIA § 103(c), commonly owned § 102(e)/(f)/(g) art cannot be used in a § 103 combination, and the same-day companion application is not prior art at all. This affects only the imicon/companion-application line of attack, not A1/B1/C1/D1/E1.

8. Bottom line

• Claim 1 and claim 4 are obvious over the combination of (A1) a known time-aligned continuous-speech recognizer [Huang/Sphinx-II] + (C1/C2) known full-text indexing/retrieval [Salton & McGill; Mauldin] + (B1) known content-based video partitioning [Zhang et al. 1993] + (D1/E1) known synchronized digital audio/video storage and delivery [Browne; DEC EP 0 633 694], and/or over F1 (US 5,136,655) if its disclosure matches its title.
• Claims 2–3, 5–6 (NLP interpreter + rule updating) obvious over the above plus the admitted known NLP processors (C2 and the patent's own admissions).
• Claims 7–10 (icons, imicons, compression) obvious over B1 + E1 and admitted-known compression.
• Claims 11–13 (Sphinx-II pipeline) obvious over A1 alone / A1 + A2 — these are close to verbatim admissions of prior art.
• Claims 14–34 cannot be assessed from the truncated text; supplying the actual claim set is required.

Overall: The patent reads as an integration of admittedly known components (ASR, IR, video segmentation, networked storage), and the specification's own background supplies both the design incentive and the "these were known and combinable" teaching. Under KSR, that profile is a strong § 103 case, weakest (relative to the rest) only in the imicon-specific claims 7–9, where the companion-application ownership/§ 103(c) issue and the coinage of "imicon" create some space for a non-obviousness argument.

Two integrity flags (as required)

1. Missing "Prior Art" section. The prior-art list you directed me to was not present in the supplied materials; I substituted verifiable references and admissions and flagged every unverified item. If a document list exists on the page you were viewing, provide it and I will re-map against those exact references.
2. Cross-section consistency. My Litigation summary earlier listed family member "KR 5 U," with a parenthetical. The Google Patents record provided here does not itself show a KR family member. No conclusion in this analysis depends on it, but the discrepancy should be noted rather than silently reconciled.