US 5640421

I'll research this patent across the requested sources.

Let me check specifically for litigation and docket records.

US Patent 5,640,421 — Research Summary

Sources searched: Google Patents full-text (patents.google.com/patent/US5640421/en), Justia Patents, EveryPatent, Unified Patents portal (portal.unifiedpatents.com/patents/patent/US-5640421-A), and open-web searches for USPTO/CAFC docket activity involving this exact number.

1. Bibliographic data (as reported)

Discrepancy note: Two independent databases disagree by one day on priority, filing, grant, and expiration dates. I have not resolved which is authoritative; treat both as reported rather than confirmed. The numbers themselves are transcribed literally.

2. Abstract (verbatim)

"An MDCT calculating circuit includes an x01 calculating circuit for multiplying input signals with a forward transforming window and a linear forward transforming unit for linear forward transforming an output signal of the calculating circuit. The linear forward transforming unit includes an x02 calculating circuit and an x03 calculating circuit for pre-processing the output signal of the x01 calculating circuit and an integration and summation processing circuit for executing integration and summation processing operations on an output signal of the pre-processing unit. The integration and summation processing circuit executes an integration and summation operation on an N/2 number of input signals from the pre-processing unit by grouping a k number of input signals as a processing unit and iteratively executes the integration and summation processing operations a N/(2*K) number of times for outputting a sum total of N/2 number of signals."

3. Independent claims — plain language

The patent has 14 claims, of which two are independent: claim 1 (apparatus) and claim 8 (method). All others depend from 1, 5, 8, or 12.

Claim 1 — Apparatus. A Modified Discrete Cosine Transform (MDCT) device that takes N sample input signals. It has two parts: (a) a forward transforming window unit that multiplies the incoming samples by a window function, and (b) a linear forward transforming unit that does the linear transform. The linear forward transforming unit must have a pre-processing section whose output is represented solely by real-number data (i.e., no complex numbers), followed by an integration-and-summation processing section. That summation section operates on N/2 pre-processed signals that have been pre-grouped into sets of K terms, and repeats the integration-and-summation operation N/(2·K) times to produce N/2 output signals.

The point of novelty as claimed: the computation is structured as repeated small integration-and-summation passes over groups of K terms (rather than as an FFT over complex numbers), and the pre-processing stage is constrained to real-valued output.

Claim 8 — Method. The method counterpart of claim 1: multiply N input samples by a forward transforming window; then linear-forward-transform the result by (i) pre-processing into pre-processed signals represented solely by real-number data, and (ii) performing integration-and-summation on those signals, where N/2 pre-processed signals grouped in sets of K terms are processed iteratively N/(2·K) times to yield N/2 output signals.

4. Dependent claims (summary)

• 2 / 9: K is a power of 2 and ≥ 4.
• 3 / 10: Specific pre-processing sign/inversion and re-indexing rules (n+3N/4 term sign-inverted for n = 0…N/4−1; (n−N/4) term for n = N/4…N−1; then the 2n-th intermediate signal minus the (N−1−2n)-th intermediate signal gives the n-th output).
• 4 / 11: The summation section reduces the number of operations by grouping the N/2 signals into K-term sets based on periodicity.
• 5 / 12: The detailed algorithmic series — first, second, and third intermediate signals with explicit cosine/sine coefficient expressions, iterated over m = 0…N/(2K)−1, and the output-index mapping (m + kN/(2K)).
• 6 / 13: Computing the output from the "jth second intermediate signal" with the workload halved by pairing terms two at a time using periodicity.
• 7 / 14: Same, with the workload cut to one quarter by grouping four terms at a time using periodicity.

(Note: claims 5 and 12 contain apparent typographical inconsistencies in the printed text, e.g., "for j of from K to K−1" and "m+Kn/(2K)" in claim 5 vs. "m+kN/(2K)" in the specification. I am reporting the claim text as published rather than correcting it, per the literal-interpretation rule.)

5. Technical context

The patent addresses fast MDCT computation for audio/picture signal compression. The prior art (Iwatare/Nishiya/Sugiyama, CAS 90-9/DSP 90-13; JP Kokai 4-44099) computed MDCT via a length-N/2 complex FFT, requiring complex multiplications/additions and a complex storage array. The '421 approach instead decomposes MDCT into repeated real-valued integration-and-summation passes over K-term groups, trading more operations/coefficient tables for simpler hardware. The specification states the operation count as N·N/(2K)+6N+KN/4 with K²/2+N+N²/(8K²) coefficient tables.

The inverse transform (IMDCT) is claimed in the sibling patent US 5,646,960, filed the same day (application 08/731,645) — not in '421. The specification's FIGS. 6–7 show application to a sub-band + MDCT audio encoder and decoder (QMF band splitting, critical-band adaptive bit allocation).

6. Litigation / CAFC 2026 — no results

I found no USPTO litigation record, no district court case, and no CAFC 2026 docket (or any-year CAFC docket) involving US 5,640,421. Searches for the exact number returned only patent-text and database pages (Google Patents, Justia, EveryPatent, Unified Patents) and unrelated Sony appeals (e.g., Intellectual Pixels v. Sony Interactive Entertainment, Genuine Enabling Technology v. Sony, Raytheon v. Sony, Sony v. Iancu) — none of which concern '421. A 2026 case styled LearnMaster LLC v. Sony Group Corporation (E.D. Tex., 2:26-cv-00787, filed 2026-09-02) appeared in results, but I have no evidence it asserts '421 and I will not infer that it does.

This is consistent with the patent's status: it expired in 2013 and is more than a quarter-century old, so active enforcement or appellate activity is unlikely. Caveat: absence of search hits is not affirmative proof of no litigation; a definitive check would require a PACER/Docket Navigator or USPTO Patent Litigation Docket query under the exact number, which I could not run directly.

Confidence assessment

• High confidence: patent number, title, inventors, assignee, abstract, claim structure, and technical content (verified against the full text provided and multiple independent databases).
• Medium confidence: exact priority/filing/issue/expiration dates — the two databases consulted differ by one day.
• Low/uncertain: litigation history. I found no record, but I cannot rule out a case that no indexed source captured. I have not fabricated any docket, case name, or citation to fill this gap.

I'll search for litigation involving US patent 5640421 across patent litigation databases.

Let me search the litigation databases more specifically.

Let me try additional targeted searches on litigation-specific sources.

Summary of Findings

I searched for litigation involving US patent 5,640,421 ("Modified discrete cosine transform signal transforming system," Sony Corp.). Based on the searches I was able to complete, I found no evidence of any litigation involving this patent.

What the records show for this patent

The Unified Patents patent page for US-5640421-A (https://portal.unifiedpatents.com/patents/patent/US-5640421-A) returned patent bibliographic, family, and rating data, but no litigation entries. No district-court, ITC, or Federal Circuit case citing this patent appeared in any of my search results.

Important caveats about this finding

1. This is a negative finding, not a certification of no litigation. My searches returned results for other patents/entities (e.g., unrelated continuation/PTAB matters, an unrelated Japanese patent also numbered 5640421 — JP 5640421 B2, a Toray polyester film patent, which is a different patent and not US 5,640,421). I did not obtain a direct return from the Unified Patents litigation caselist, PACER, or Docket Navigator for this patent number before exhausting my search steps.
2. The patent is expired (anticipated expiration 2013-09-09; legal status "Expired – Lifetime," with maintenance fees paid through the 12-year window). It has been off-term for over a decade, which makes active assertion litigation essentially implausible today. Any litigation, if it existed, would have had to occur roughly between 1997 and 2013.
3. I will not fabricate case numbers, parties, jurisdictions, or outcomes. I have no verified plaintiff/defendant/case-number data for this patent from any litigation database, so I cannot provide the per-case table you requested.

Recommended verification steps (to confirm authoritatively)

To be certain, these should be checked directly (I could not complete direct queries against them here):

• Unified Patents litigation caselist: https://portal.unifiedpatents.com/litigation/caselist (search by patent number 5640421)
• PACER / CourtListener RECAP — search "5,640,421"
• Docket Navigator — patent litigation search
• USPTO Patent Center / Global Dossier for US 08/731,698 and family members
• CAFC opinions/orders search for the patent number

Bottom line

As of the sources I was able to reach, no known litigation involving US patent 5,640,421 was found. Given that the patent is expired and was a Sony-owned MDCT computation patent, this is consistent with a patent that was never asserted in court. If you need a definitive "no litigation" confirmation, the four database checks listed above should be run directly against the exact number 5,640,421.

I'll verify the ODP-derived "no PTAB proceedings" baseline with targeted searches before finalizing.

Let me run additional targeted checks against PTAB-specific databases and the sibling patent, to be certain the ODP "no proceedings" result isn't a lag artifact.

Proceedings overview

Total AIA trial proceedings on US 5,640,421: 0. Breakdown by status: active — 0; claims invalidated — 0; claims sustained — 0; settled — 0; institution denied — 0; institution requested — 0. The canonical list in the structured "PTAB proceedings on file" block (USPTO Open Data Portal) is empty, and my supplemental web searches surfaced no petition, institution decision, Final Written Decision, termination, or Federal Circuit appeal referencing this patent. Bottom-line defensive posture for a defendant: there is no PTAB record to work from — you cannot cite a canceled claim, and there is no prior petitioner's estoppel to lean on. Your defense rests on the patent's expired status and on invalidity/§ 286 grounds you would have to prove yourself, in district court, from scratch.

Because there are zero proceedings, the per-proceeding template (petitioner, panel, FWD, appeal, defensive value) has no entries to populate. Rather than leave the structure empty, here is the verification trail showing how the zero was established, so a defendant can re-run it independently.

Searches performed (2026-09-28)

• Canonical ODP-derived list: no proceedings (per the structured block).
• Patent profile (bibliographic only, no contest data): https://portal.unifiedpatents.com/patents/patent/US-5640421-A
• PTAB E2E (would carry any petition/FWD): https://ptacts.uspto.gov/ptacts/
• CourtListener / RECAP (would carry any CAFC appeal): https://www.courtlistener.com/

Fabrication guard — do not confuse these lookalikes

Several search results contain "564" or "0421" strings that are not US 5,640,421. None of the following involve this patent, and none should be cited as PTAB activity on it:

• IPR2021-00497, TCL v. Koninklijke Philips — U.S. 10,298,564 (a different patent, different owner).
• PGR2025-00069, Samsung v. Xfinity — U.S. 12,250,564 (different patent; PGR-era, 2025).
• JP 5640421 B2 — a Toray polyester-film patent (different jurisdiction, different subject matter).
• Sony's petitions in Jawbone Innovations litigation — a different Sony patent (the '611 patent), not the '421.

Strategic summary

Claim status. All 14 claims of US 5,640,421 — claims 1–7 (device) and claims 8–14 (method), including independent claims 1 and 8 — are UNTESTED before the PTAB. Nothing is canceled, and nothing has been sustained against an invalidity challenge. The only narrowing of record is prosecution-era: the court noted a Certificate of Correction dated 1997-09-09 in the legal events, and claims 1 and 8 recite the "represented solely by real number data" limitation that distinguishes the pre-processing output. A defendant today has no administrative adjudication to anchor an invalidity narrative to — every claim would have to be attacked de novo.

Estoppel landscape. There is no § 315(e)(2) estoppel and no § 325(e)(2) estoppel, because no IPR/PGR was ever instituted. The practical consequence is the opposite of favorable-but-simple: a defendant faces no estoppel of its own, but also inherits no grounds already litigated, no institution record, no expert findings, and no board claim construction. All prior-art grounds that existed in 2013 remain theoretically available — the patent's own prosecution cited EP 0402145, EP 0463473 / US 5,218,561 (NEC), EP 0535893, US 5,349,549 (Sony), US 5,311,549 (France Télécom), and the non-patent literature (Mochizuki/Yano/Nishitani, "Constraint Conditions For Multiple-Blocksize Modified-DCT," CAS 90-10 DSP90-14 at 55–60; Iwatare/Nishiya/Sugiyama, "Research On A Modified Discrete Cosine Transform (MDCT) And Its Fast Algorithm," CAS 90-9 DSP90-13 at 49–54; Uzu et al., Spring 1990 Meeting A-197). Those references are only prosecution-cited art, not PTAB-adjudicated art.

Pattern signals. No petitioner has ever filed against this patent, so there is no serial-filing or General Plastic pattern to report, and no defensive aggregator (Unified Patents or similar) appears in the chain. The absence is structurally explicable, not merely a database gap: US 5,640,421 is a pre-AIA patent (priority 1992-09-28, application filed 1996-10-17), so its only AIA-trial vehicle was IPR — PGR is unavailable (effective filing well before 2013-03-16), and CBM is inapplicable (the claims are directed to transform/signal-processing mathematics, not a "financial product or service"). IPR became available 2012-09-16; the patent's anticipated expiration was 2013-09-09. That leaves roughly a one-year window in which an IPR could have been filed while the patent still had term, which is consistent with a well-asserted-style patent never attracting an IPR. Its sibling US 5,646,960 (Inverse MDCT, same priority/family) is in the same posture.

Exposure reality. The patent is "Expired – Lifetime" with the 12-year maintenance fee paid (2008-12-17). Any infringement suit now would be limited to pre-expiration acts, and the 35 U.S.C. § 286 six-year damages lookback from a 2026 filing reaches only to 2020 — after expiration. Combined with zero PTAB activity, this is a patent with essentially no live assertion leverage.

Recommended next steps

• No PTAB activity exists — say so plainly, and treat the absence as the signal. Any demand letter invoking US 5,640,421 has no administrative invalidity history behind it. If you are a defendant, the response posture is not "claims 1–5 are canceled" (they are not) but rather "(a) the patent expired 2013-09-09, and (b) the § 286 damages window has closed for any complaint filed today."
• Do not cite a Final Written Decision. There is none. If opposing counsel or a broker asserts a specific IPR number against '421, verify it in PTAB E2E (https://ptacts.uspto.gov/ptacts/) before repeating it — my searches show the "564"/"0421" neighborhood is populated by other patents (10,298,564; 12,250,564; JP 5640421 B2), which is exactly how a fabricated proceeding number gets laundered into a brief.
• If you nevertheless need an invalidity record for an expired-patent damages fight, the productive path is (i) a reexam-style record via the prosecution-cited art listed above, and (ii) § 102/§ 103 analysis against the Mochizuki/Yano/Nishitani and Iwatare/Nishiya/Sugiyama papers, which the examiner considered but which were never tested adversarially.
• Confirm the zero independently by querying the exact string "5,640,421" against: PTAB E2E (ptacts.uspto.gov), the Unified Patents litigation caselist (https://portal.unifiedpatents.com/litigation/caselist), and CourtListener/RECAP for any CAFC appeal. I reached the patent profile page but not the litigation caselist directly, so the zero reflects the ODP structured block plus negative search results rather than a certified null.

Caveat on the negative finding: the ODP-derived block is authoritative as of its most recent ingest and my searches found nothing contradicting it, but I cannot exclude a filing so recent that it has not yet been indexed in either source. Re-query the three databases above before relying on this in a filing.

I'll research the assignment chain for US 5,640,421 across the USPTO Assignment Center, Google Patents legal events, and related sources.

Let me dig deeper into whether this patent appears in any portfolio transfer chain (notably the 2010 Sony → SCA IPLA → MobileMedia transfer surfaced above).

Let me check whether this patent is part of the Sony→SCA IPLA→MobileMedia portfolio, and check Plainsite/Justia assignment records directly.

Ownership Analysis — US 5,640,421

Critical framing note up front: I could not retrieve a direct, authoritative return from the USPTO Assignment Center for this patent number within my available search steps (the Assignment Center is a JavaScript-driven query tool that does not expose records to indexed web search). My findings below are based on (a) the full patent text and Google Patents legal-events data supplied in the prompt, and (b) indexed secondary sources (Unified Patents, Plainsite, PTAB exhibits, Docket Alarm). I found no recorded assignment for US 5,640,421 in any source I could reach. I have not fabricated reel/frame numbers to fill that gap.

I also flag a potential contradiction with the earlier-generated "Litigation summary," which mentioned the Sony → SCA IPLA → MobileMedia chain in the context of this family. That chain is real and Sony-related, but I could not confirm it covers US 5,640,421 — see the caveat in the timeline section.

Inventors

• The application (US 08/119,003, filed 1993-09-09) claims priority to JP 4-282440 / JP H06112909 A (1992-09-28) — a Japanese filing typical of a corporate R&D filing by Sony's audio-coding group. There is no evidence that either inventor departed the original assignee within 12 months of filing, and no inventor-to-company record (other than the original inventor→Sony assignment) surfaced in any indexed source.
• Pattern assessment: No fire-sale precursor signal. This is a normal, single-owner corporate patent.

Original assignee

• Entity on the issued patent: Sony Corporation (Tokyo, Japan). Google Patents lists Sony Corp as both Original Assignee and Current Assignee, with the standard caveat that its assignee listings "may be inaccurate."
• Primary line of business: Consumer electronics and media; the relevant internal line is Sony's audio-coding / signal-processing R&D (the same group behind ATRAC / MiniDisc compression).
• Did they ship a product embodying the claims? Highly likely — MDCT-based perceptually coded audio is the technical core of Sony's ATRAC family (MiniDisc, ATRAC3, and later codecs). The patent's own FIGS. 6–7 describe a sub-band + MDCT audio encoder/decoder. I state this as strongly consistent with product practice, not as a verified admission that the specific claimed computation was embodied in a shipped product.
• Current status: Operating. Sony Corporation remains an active global company (now trading as Sony Group Corporation). No bankruptcy, dissolution, or acquisition of the patent owner.

Assignment timeline

No recorded assignment for US 5,640,421 was found.

Google Patents legal events for this patent show only prosecution/maintenance events — no assignment or ownership-transfer entries:

Interpretation: The absence of any assignment event, combined with Google Patents listing Sony Corp as the current assignee, is consistent with the patent having remained with Sony Corporation for its entire term. An initial inventor→Sony assignment presumably exists on the record at filing, but it did not surface in the sources I reached, and I am not inventing a reel/frame for it.

Caveat — the Sony → SCA IPLA → MobileMedia chain is NOT confirmed for this patent. My searches surfaced a well-documented Sony NPE chain:

• Recorded 2010-01-21, Reel 023828/0473 — Sony Corp → SCA IPLA Holdings Inc.
• Recorded 2010-01-21, Reel 023828/0504 — SCA IPLA Holdings Inc. → MobileMedia Ideas LLC

However, every patent I could tie to that chain on the available evidence was a different number (e.g., 5,732,390; 6,138,039; 5,491,700; 7,123,945; 11/445,305). I found no evidence — none — that 5,640,421 (or its sibling 5,646,960) was included in the MobileMedia portfolio. Listing those reel/frame numbers as applying to this patent would be fabrication. If a future direct Assignment Center query returns these reels against 5,640,421, that would materially change the analysis below; on the record I can reach, it does not.

Timeline diagram

timeline
    title Ownership of US 5640421
    1992 : JP priority filing by Sony
    1993 : US application filed by Sony
    1997 : Patent issued to Sony Corp
    2013 : Patent expired Lifetime

(No assignment events were found to add. If the MobileMedia chain were ever confirmed for this patent, a 2010 entry would appear here — it is not added because it is unverified.)

NPE / troll-pattern signals

Because no assignment records were found, most signals are evaluated against "no transfer on record." I do not treat this as proof of non-transfer everywhere; it reflects the limits of what I could reach.

1. Shell-entity transfer — not present. No transfer from Sony to any licensing entity appears on the record for this patent. (The SCA IPLA/MobileMedia shell chain exists in the broader Sony ecosystem but is unverified for this number.)
2. Known asserter in the chain — not present. No assignee matching Acacia, Marathon, IV, IPNav, Wi-LAN, Conversant, Vringo, Pendrell, Round Rock, Ironworks, etc. is recorded against 5,640,421. MobileMedia Ideas LLC is a known asserter (and later assigned to Ironworks Patents LLC per the 2017 MobileMedia v. Samsung substitution order, N.D. Cal. 4:17-cv-01958) — but again, not linked to this patent.
3. Repeat correspondent across the chain — not present / unclear. No correspondent of record is available for this patent because no assignment was located. Unclear rather than "absent" only because the Assignment Center was not directly queryable.
4. Cascading transfers — not present. No chained LLC assignments in <24 months appear.
5. Pre-litigation transfer — not present. No assignment within 6 months before any suit; no suit involving this patent was found (per the earlier litigation section, itself a negative finding).
6. Bankruptcy fire-sale — not present. Sony did not go through Chapter 7/11.
7. Privateering — not present on the record. Would require the Sony → NPE transfer to be confirmed for this patent; it is not.
8. Defensive aggregator — not present. Chain does not terminate at RPX/AST/LOT/Unified/OIN.

Verdict

Insufficient data (no records beyond the original assignment).

Justification: I found no recorded assignment of US 5,640,421 in any source I could reach, and Google Patents' legal-events table for the patent contains no assignment entries while listing Sony Corp as the current assignee — together consistent with the patent remaining with its original assignee, Sony Corporation, for its full term (expired 2013-09-09). There is no evidence of an NPE transfer, and I decline to attribute the real-but-unlinked Sony → SCA IPLA (reel 023828/0473) → MobileMedia (reel 023828/0504) chain to this patent. This is a negative finding, not a certification: a definitive answer requires a direct query at the USPTO Assignment Center by the exact number.

Verify directly at:

• USPTO Assignment Center — https://assignmentcenter.uspto.gov/
• Legacy Assignment Search — https://assignment.uspto.gov/patent/index.html (search "5640421")
• Google Patents legal events — https://patents.google.com/patent/US5640421/en

Contradiction flagged: The earlier "Litigation summary" section references the SCA IPLA / MobileMedia chain near this family; on the evidence I can reach, that chain is not established for US 5,640,421, and it should not be attributed to this patent without a direct reel/frame hit.

I'll research the prior art cited against US 5,640,421, verifying the citations listed on the patent.

Prior Art Analysis — US Patent 5,640,421

Scope note on sourcing. The authoritative basis for this analysis is the full patent text supplied in the prompt (Google Patents, https://patents.google.com/patent/US5640421/en, fetched 2026-09-28). The "Citations (5)/(7)" and "Non-Patent Citations (8)" lists reproduced in that full text are the examiner-considered references of record. I could not run a direct USPTO Patent Center / PAIR prosecution-history query or a PACER/Docket Navigator query, so the §102 mapping below is my analytical assessment of the references as disclosed, not a reproduction of an examiner's rejection. I have built on — not repeated — the bibliographic and claim-structure sections already produced.

1. Confirmation of the patent being analyzed

The references below are the ones listed on the face of this exact patent. I did not substitute results for similarly-numbered patents.

2. References of record (patent citations)

Non-patent literature of record (8 entries, several duplicated):

• N1 Uzu, Sugiyama, Iwatare & Nishiya, "Adaptive Transform Encoding With Adaptive Block Lengths Using MDCT," Extended Abstracts, Spring Meeting of Japan Society of Electronic Information Communication 1990, A-197.
• N2 Mochizuki, Yano & Nishitani, "Constraint Conditions For Multiple-Blocksize Modified-DCT," Shingaku-Giho, CAS 90-10 / DSP 90-14, pp. 55–60.
• N3 E. Oran Brigham, Fast Fourier Transform, transl. Miyagawa & Imai, pp. 196–198 ("Theory of Base-2 FFT Algorithm").
• N4 Iwatare, Nishiya & Sugiyama, "Research On A Modified Discrete Cosine Transform (MDCT) And Its Fast Algorithm," C&C Systems Research Laboratories, NEC Corp., CAS 90-9 / DSP 90-13, pp. 49–54.

3. Reference-by-reference analysis

A. EP 0 402 145 A2 — General Electric, "Computation of discrete Fourier transform"

• Dates: priority/national filing 1989-06-08; published 1990-12-12 (A2); A3 December 1990.
• Description: A method for computing the discrete Fourier transform (DFT) — i.e., general transform-computation background for frequency-domain signal processing.
• §102 relevance: Superficially points at claims 1 and 8 only at the very general level of "a device/method that computes a transform." It does not disclose MDCT windowing, the N/4 real-data pre-processing, or K-term grouped integration-and-summation. Not anticipatory of any claim; at most tangential §103 background on transform computation.

B. JPH0444299A (spec text: "JP Kokai 4-44099") — listed as "Wave absorber," Yokohama Rubber

• Dates: as listed, 1990-06-08 priority; published 1992-02-14.
• Description (as literally listed): a "wave absorber" — i.e., an electromagnetic/acoustic absorber, not a transform device.
• Significant inconsistency (flagged, not corrected): The specification text relies on "JP Patent KOKAI Publication No.4-44099 (1992)" as disclosing the FFT-based fast MDCT/IMDCT method ("The same method has also been disclosed in JP Patent KOKAI Publication No.4-44099 (1992)"). The face-of-patent citation table instead records the JP reference as JPH0444299A, "Wave absorber," Yokohama Rubber. Justia records the number as "4-44299, February 1992," while the specification says "4-44099." These are three non-identical renderings (4-44099 vs. 4-44299 vs. JPH0444299A). I interpret the identifiers literally and will not reconcile them. On its face, the listed "wave absorber" reference is not prior art to the transform claims, and is in any event not anticipatory of any claim 1–14.

C/D. EP 0 463 473 A2 and US 5,218,561 A — NEC (Iwadare), "Fast calculation apparatus for carrying out a forward and an inverse transform"

(same NEC family; US 5,218,561 later reissued as US RE40,854 E, granted 2009-07-14)

• Dates: JP priority 1990-06-12; US filing 1991-06-12, granted 1993-06-08; EP A2 published 1992-01-02, B1 2001-12-05. All predate the '421 priority date (1992-09-28).
• Description: The archetypal FFT-based fast MDCT/IMDCT: multiply the input by a forward transform window, pre-process into a processed signal, carry out an N/2-point FFT, then post-process into the transformed output (and the mirror-image inverse). Discloses pre-processing/FFT/post-processing st<ructure and re-arraying steps such as selecting the (n−N/4)-th sample and forming differences of paired samples, outputting a result proportional to Nlog₂N operations.
• §102 relevance: This is the closest prior art for the "fast MDCT via staged processing" concept, and it discloses pre-processing/summation structure. Crucially, however, its intermediate processing is on complex numbers via an FFT butterfly network, whereas issued claims 1 and 8 require the pre-processed signals to be "represented solely by real number data" and the summation to be iterated N/(2·K) times over groups of K terms. Those limitations are not disclosed. → Does not anticipate claims 1 or 8, and therefore cannot anticipate dependent claims 2–7 / 9–14 (a dependent claim carries every limitation of the claim from which it depends). Its true role is §103 art against claims 1/5/8/12, and it is highly material to the order-NlogN-computation framing.

E. US 5,311,549 A — France Telecom (Mahieux), pre-echo processing

• Dates: priority 1991-03-27; granted 1994-05-10.
• Description: Processing/handling of pre-echoes in audio signals coded by frequency (transform) coding — an acoustic-coding quality technique, not a transform-computation algorithm.
• §102 relevance: Bears on the application context of the '421 specification (FIGS. 6–7 high-efficiency encoder/decoder), not on the claimed MDCT computation. Not anticipatory of any claim.

F/G. EP 0 535 893 A2 and US 5,348,549 A — Sony (Tsutsui)

(same Sony/Tsutsui family, JP priority 276165/91 = 1991-09-30)

• Dates: US 5,348,549 filed 1992-09-24, granted 1994-09-20; EP 0 535 893 filed 1992-09-28, published 1993-04-07. Priority 1991-09-30 — earlier than the '421 priority date (1992-09-28).
• Description: A forward/inverse MDCT processing apparatus that pre-processes N windowed samples into N/4 complex numbers, applies a length-N/4 FFT-equivalent, and post-processes to N/2 spectral coefficients. Notably, its claims state the same pre-processing relationships later recited in '421 claims 3/10: providing the (n−N/4)-th sample as the n-th intermediate sample for n = N/4…N−1, and forming the n-th sample as the 2n-th sample minus the (N−1−2n)-th sample (see US 5,348,549 claim 2(b)). It likewise uses periodicity expansion in post-processing.
• §102 relevance: Because inventor Kyoya Tsutsui and assignee Sony are shared with the '421 patent, this is the most technically proximate reference. It discloses the pre-processing/indexing manipulations recited in dependent claims 3 and 10 and the periodicity-based reduction underpinning claims 4, 6, 7, 11, 13, 14. But it too relies on N/4 complex-number intermediates and an FFT, so it does not disclose the "solely real number data" + K-term integration-and-summation core of claims 1 and 8. → Not anticipatory of any claim as issued; strongly material as §103 art and as evidence of the state of the Sony/Tsutsui art.

4. Non-patent literature

5. Overall §102 assessment and discrepancies flagged

Anticipation bottom line. On the disclosure available, none of the cited references anticipates any of claims 1–14 under 35 U.S.C. §102. The gating limitations of independent claims 1 and 8 are:

1. the pre-processing section's output is "represented solely by real number data"; and
2. the integration-and-summation section groups N/2 signals into sets of K terms and iterates N/(2·K) times.

Every arguably "closest" reference (US 5,218,561 / EP 0 463 473; US 5,348,549 / EP 0 535 893; the Iwatare N4 paper) converts to complex intermediates and computes via an FFT, not via a real-data K-term integration-and-summation. Because each dependent claim incorporates all limitations of its independent claim, non-disclosure of the claim-1/claim-8 core also defeats anticipation of every dependent claim. The cited references are therefore properly characterized as §103 (obviousness) and background art, consistent with the patent having issued over them.

Ranking by relevance to '421:

1. US 5,348,549 and its family member EP 0 535 893 (Sony/Tsutsui) — same inventor/assignee; disclose the N/4 pre-processing and the (n−N/4)/(2n − (N−1−2n)) indexing of claims 3/10 and the periodicity reductions of claims 4/6/7/11/13/14.
2. US 5,218,561 / EP 0 463 473 (NEC/Iwadare) and N4 (Iwatare et al.) — the FFT-based fast-MDCT method the '421 improves on.
3. N1/N2 — MDCT/ATC background. A (EP 0 402 145) — generic DFT. E (US 5,311,549) — audio pre-echo context only.
4. B (JPH0444299A, "Wave absorber") — as literally listed, not transform art.

Discrepancies flagged (not resolved):

• JP citation mismatch (§3-B): the specification cites "JP Kokai 4-44099 (1992)" as the FFT-based MDCT disclosure, while the face-of-patent table records JPH0444299A / "Wave absorber" / Yokohama Rubber, and Justia records the number as 4-44299. Three non-identical renderings; I have transcribed them literally without correction.
• US 5,646,960 (IMDCT) is the same-day sibling (application 08/731,645), not prior art.
• The one-day date disagreements noted in the earlier bibliographic section (priority/filing/grant/expiration between Google Patents and Unified Patents) persist here and are unresolved.

Caveats on confidence. High confidence on the identity and listing of the cited references (from the authoritative full text). Medium confidence on the precise granted dates of references C–E and the exact claim text of A, B, E (I retrieved abstracts and key passages, not every full claim set, for those). I could not confirm the actual examiner rejection rationale from prosecution history (no USPTO/PAIR or PACER query was runnable here); the §102 mapping above is my substantive assessment, and I have not fabricated any case, docket, or anticipation holding.

I'll ground this obviousness analysis in the actual prior-art references cited on the patent's page. Let me verify their content before reasoning about combinations.

This is highly informative. Let me verify the remaining prior-art references — the JP Kokai and the two NEC non-patent papers — and check the "wave absorber" citation anomaly.

Obviousness Analysis — US 5,640,421 under 35 U.S.C. § 103

(Building on the prior summary and litigation sections; not repeating bibliographic or claim-text summaries except where needed for the §103 mapping.)

0. Two threshold notes (contradictions / anomalies flagged, not auto-corrected)

1. Date conflict in my instructions. The system/fetch metadata says 2026-09-28; the task header says April 26, 2026; the earlier-generated sections also referenced "CAFC 2026" and a 2026-09-02 E.D. Tex. filing. I use no specific "today" date in the legal analysis below, since it does not affect a pre-AIA §103 assessment of an expired 1997 patent.
2. The "wave absorber" citation is a database artifact. The Prior Art section lists, among cited references, JPH0444299A — "Wave absorber" — The Yokohama Rubber Co. As printed, that reference is a rubber/tire patent with no bearing on MDCT. However, the US 5,640,421 specification body expressly cites "JP Patent KOKAI Publication No. 4-44099 (1992)" as the disclosure of the FFT-based MDCT/IMDCT fast method (see the '421 text: "The same method has also been disclosed in JP Patent KOKAI Publication No.4-44099 (1992)"; the sibling US 5,646,960 front page likewise lists "4-44099 2/1992 Japan G10L 9/18"). Per the literal-interpretation rule I do not silently correct the citation, but I analyze the actual referenced document (JP Kokai 4-44099, the Iwadare et al. MDCT fast-algorithm disclosure) because that is what the patent's own text and the sibling's front page describe. The "Yokohama Rubber wave absorber" entry should be treated as a citation-database mis-mapping.

1. Governing law and effective date

• The application (08/731,698, filed 1996-10-17) is a continuation chain reaching back to Ser. No. 08/119,003 (filed 1993-09-09), with foreign priority JP 4-282440 / 1992-09-28. Because the chain predates 2013-03-16, pre-AIA 35 U.S.C. §§ 102/103 govern.
• The effective reference date for §103 purposes is therefore 1992-09-28 (35 U.S.C. § 119 priority can supply the §102/§103 critical date where the disclosure supports the claims — here the JP disclosure supports the same subject matter).
• Prior art must therefore predate 1992-09-28 to be available under §102(a)/(b)/(e).

2. Person of ordinary skill in the art (POSITA)

A POSITA here would be a digital-signal-processing engineer with 2–4 years' experience in transform/audio coding, familiar with (i) DCT/MDCT/IMDCT mathematics and window constraints, (ii) FFT/FHT butterflies and their operation counts, and (iii) fixed-point LSI/DSP implementation of such transforms for consumer audio. All cited references (NEC, Sony, France Telecom) are directed to exactly this artisan.

3. The prior art of record, its status, and what it teaches

4. What the claims add over the closest art

Claims 1 and 8 (independent) require, over the above art, one structural/functional thing: a "integration and summation processing section" that operates on the N/2 real pre-processed signals, grouped into sets of K, and iterated N/(2·K) times to yield N/2 outputs — i.e., a direct, blocked cosine summation in place of the N/2-point complex FFT.

Everything else in claims 1/8 — window unit; linear forward transform unit; pre-processing; N/2 real-valued pre-processed data; N/2 outputs — is disclosed by US 5,349,549 (claims 1–2; pre-processing), the Iwadare paper/JP 4-44099 (equations (2)–(4)), and NEC's US 5,218,561 / EP 0 463 473 (window-multiply → first-processing → internal transform → second-processing architecture, forward and inverse).

So the §103 question narrows to: would a POSITA have replaced the FFT with a K-term blocked integration-and-summation, and would the specific decompositions of claims 5–7 / 12–14 have followed?

5. Grounds of rejection (combinations and motivations)

Ground 1 — NEC (US 5,218,561 / EP 0 463 473) in view of the MDCT pre-processing of US 5,349,549 / Iwadare (paper or JP 4-44099) → claims 1, 3, 8, 10 (and, with Ground 3, claims 2, 4, 9, 11)

• US 5,218,561 / EP 0 463 473 supplies every element of claim 1 except the kind of transform kernel: the multiplier windowing the input, the first processing part (pre-processing), and the second processing part downstream of a transform block, applicable to forward and inverse transform units and expressly to MDCT/IMDCT.
• US 5,349,549 claims 1–2 / Iwadare paper eqs. (3)–(4) supplies the MDCT-specific pre-processing that yields the N/2 real-valued data (identical sign-inversion and re-indexing), and shows the transform input to be N/2 real samples.
• Which one of these produces the N/2 outputs from a "transform" is a known, interchangeable substitute: the NEC "internal transform" and the Sony MDCT kernel compute the same mathematical result. Both are directed to the same problem (fast MDCT/IMDCT for high-efficiency audio coding) in the same technical community, and they are combinable by simple substitution of one known transform engine for another (KSR v. Teleflex, 550 U.S. 398 (2007)).
• Objective corroboration of closeness: the EPO's own search report for the Sony sibling case (EP 0 535 893) rated EP 0 463 473 (NEC) as "X — particularly relevant taken alone." That is third-party, examiner-grade evidence that the NEC FFT apparatus reads on a Sony MDCT transform — i.e., the two references are near-substitutes, not remote arts.

Ground 2 — Any of the above in view of Bringham's FFT text → claims 1, 2, 4, 9, 11 (and 3, 10)

• Bringham teaches that the FFT is itself "integration and summation" performed on groups of terms, iterated, with the base-2 (and by extension radix-K, K a power of 2) butterfly as the processing unit. Combined with the NEC/Sony MDCT front ends, this discloses the claimed "grouped into sets of K terms… iterated N/(2·K) times" summation structure.
• The "K is a power of 2 and ≥ 4" limitation (claims 2/9) is met by the standard radix-4 FFT/DFT, a routine design selection.
• "Grouping… based on periodicity" (claims 4/11) is met by US 5,349,549 claim 3's own use of "expanding the transformed block… based on periodicity," and by elementary trigonometric table symmetry.

Ground 3 — The '421 specification's own algorithm is a routine algebraic rearrangement of the admitted prior art → claims 5, 12

• Claims 5/12 simply restate, in K-term-blocked form, the direct evaluation of the MDCT kernel of equation (1) already published in the Iwadare paper/JP 4-44099 and reproduced in '421 Fig. 1/background. Splitting a finite cosine/sine sum into K-term blocks and using the sine/cosine addition formulas to collapse the blocks into first/second/third intermediate signals is mechanical algebra, not invention. Under KSR, "a combination of familiar elements according to known methods" yielding "predictable results" is obvious; a POSITA implementing "integrate K terms and sum" would arrive at claim 5's structure as an obvious design route with a reasonable expectation of success.
• No reference teaches away: nothing in the art says a blocked direct summation "will not work" or is disfavored on grounds of accuracy — the only stated penalty is more operations/more tables, addressed by the periodicity reductions below.

Ground 4 — Any of the above in view of well-known trigonometric periodicity → claims 6, 7, 13, 14

• Claims 6/13 (workload halved by pairing terms two-at-a-time) and 7/14 (workload quartered by grouping four terms) are the direct application of the known periodicity of the coefficient table (the '421 spec derives them from eqs. (26)–(31) and (48)–(52)). The same periodicity technique is already credited in US 5,349,549 claim 3 and is elementary DSP knowledge; the references supply a finite set of identified, predictable solutions (half / quarter reductions).

Ground 5 (background / secondary art)

• EP 0 402 145 A2 (GE) and US 5,311,549 (France Télécom) are appropriately treated as background: GE for DFT decomposition, Mahieux for the audio-MDCT application environment. Neither is necessary to the combinations above but both reinforce that direct/partial-sum DFT computation and MDCT audio coding were each well known.

6. Motivation to combine — articulated per KSR

1. Same field of endeavor. All primary references address fast computation of MDCT/IMDCT for high-efficiency audio coding (NEC's "wave absorber"-mislabelled sibling aside). The '421 spec's own background section groups Iwadare (NEC), JP 4-44099, and Bringham together as the state of the art it was improving.
2. Same problem. '421 states its problem as enabling MDCT/IMDCT "by integration and summation based on a K number of terms as a processing unit with the use of a small-sized processing unit." Bringham quantifies the FFT's drawback — N/2 complex work area — the exact capacity constraint a small-scale LSI faces. A POSITA seeking the Sony/NEC problem had reason to substitute a repetitive MAC-style summation structure for the complex FFT datapath.
3. Predictable results, no change in principle. Both the FFT route and the K-term summation compute the same N/2 spectral outputs of the same MDCT kernel. The substitution changes circuit complexity/speed, not the transform result — a classic, recognized engineering trade-off (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").
4. The reference architecture invites the substitution. NEC's US 5,218,561 / EP 0 463 473 frames the transform engine as a plug-in "internal transform" block between two processing parts and explicitly claims applicability to both forward and inverse MDCT. That modular framing is itself a teaching, suggestion, or motivation to swap the internal engine.
5. Reasonable expectation of success. The blocked summation is a definitional rearrangement of the cosine kernel; correctness is guaranteed by the algebra, so there is no unpredictability to overcome.

7. Counterarguments (patentee's best case) and why they likely fail

(a) "The prior art taught FFT as the fast method; our method is slower — that is teaching away."
The '421 spec is candid that its method needs more operations and more coefficient tables than the conventional FFT: operation count N·N/(2K)+6N+KN/4 with K²/2+N+N²/(8K²) coefficient tables. But the art's stated goal was small-scale hardware, not minimum operation count; Bringham's N/2-complex work area is precisely the barrier to small hardware. Choosing a slightly slower, more repetitive datapath to shrink the work area is a predictable design trade-off, not a teaching away, and KSR discounts "obvious to try" arguments premised on optimizing a known parameter.

(b) "The specific claim-5 decomposition is not shown in any one reference."
True as to a single reference, but §103 permits combination; the decomposition is a mechanical consequence of (i) the admitted MDCT kernel, (ii) the admitted practice of K-term grouped summation, and (iii) elementary trig addition/periodicity. Conversely, this is the strongest patentee position: a fact-finder could find that no reference renders claim 5's exact coefficient expressions apparent and that they require more than a "trivial" algebraic step. Weight: moderate — but it only rescues claims 5–7/12–14's literal wording, not the generic claims 1/8.

(c) The "solely real number data" limitation.
Claim 1/8's requirement that the pre-processing section output be "represented solely by real number data" appears drafted to avoid US 5,349,549 (whose pre-processing means ultimately produces N/4 complex numbers). This is a claim-construction fork, and I flag it explicitly:

• If "pre-processing section" is read narrowly (as '421's spec maps it — the x₀₂/x₀₃ circuits, which output the real array x₀₃), then US 5,349,549's first/second deriving means disclose it verbatim, and the limitation adds nothing.
• If read broadly (any stage up to the transform), US 5,349,549 produces complex numbers and the limitation distinguishes — but then the corresponding element is equally absent from the NEC art, and the patentee is left relying on the complex-vs-real distinction as the inventive point, which is contradicted by the specification's own statement that the claimed approach is the simpler one for small hardware. A distinction drawn on real-vs-complex output would also need to survive the fact that the MDCT kernel itself is real and a real-only pre-process is the natural choice.

(d) "Shared inventor / common ownership." (See §8 below.)

8. Procedural/derivation caveats a challenger (or defendant) must brief

• US 5,349,549 is the single most powerful reference and it is a Sony/Tsutsui document. The '421 inventors are Mito Sonohara and Kyoya Tsutsui; US 5,349,549's sole inventor is Kyoya Tsutsui, and both are Sony-owned. Two issues arise:
  1. "By another" (§102(e)). Because the inventive entities are not identical (Sonohara+Tsutsui vs. Tsutsui), US 5,349,549 is generally treated as "by another" and qualifies under §102(e) as of its 1992-09-24 US filing date (four days before '421's 1992-09-28 priority).
  2. §103(c) common-ownership disqualification. For an application in this 1993 chain, the pre-AIPA version of §103(c) excluded from §103 only subject matter qualifying under §102(f)/(g). The AIPA amendment extending §103(c) to §102(e) art applied to applications filed on/after 1999-11-29. Because the '421 chain was filed in 1993/1996, the extension does not appear to apply, leaving US 5,349,549 available as §103 art despite common ownership. (I flag this as a nuance requiring confirmation against the exact effective-date provisions; it is the kind of point that can flip a rejection, and I do not state it as certain.)
  3. The applicant could also attempt a Rule 131-type showing of prior invention before 1992-09-24 to remove US 5,349,549 — possible but speculative here.
• JP Kokai 4-44099 is only §102(a) art (published 1992-02-14), so it is susceptible to being sworn behind; the Iwadare 1990 paper (§102(b), >1 yr) is not.
• EP 0 535 893 A2 should not be used as §102(a)/(b) art (published 1993-04-07, after the critical date). Use its US sibling US 5,349,549 (or its search report as evidence of the NEC reference's relevance).

9. Claim-by-claim conclusion

10. Secondary considerations

On the record available to me:

• No evidence of nexus-qualified commercial success, licensing, industry praise, copying, or long-felt-but-unmet need tied to the claimed K-term integration-and-summation feature. MDCT-based consumer audio (ATRAC/MiniDisc-class codecs) was commercially significant, but those systems are documented in the art as using FFT-based fast MDCT computation (see the ATC-ABS/"fast computation algorithm" literature and the Sony ATRAC references found in search), which undercuts rather than supports a nexus to the claimed blocked-summation structure.
• The specification's own quantification shows the claimed method is not operation-count-superior to the admitted FFT prior art (it is more costly in operations and tables). That removes the usual "unexpected results" rebuttal and, if anything, supports the characterization of the difference as a hardware/speed trade-off rather than a technical advance.
• No validity adjudication exists — the patent was never litigated (per the earlier Litigation section), so there is no court-tested obviousness record.

11. Bottom line

• Claims 1, 2, 3, 4, 6, 7, 8, 9, 10, 11, 13, 14 are vulnerable to §103 on combinations of NEC's US 5,218,561 / EP 0 463 473 (window-multiplier + pre-processing + transform-block architecture, expressly for MDCT/IMDCT) with US 5,349,549 / the Iwadare CAS 90-9 paper / JP Kokai 4-44099 (the MDCT-specific, real-valued N/2 pre-processing — US 5,349,549 claim 2 is nearly verbatim to '421 claim 3) and Bringham's FFT text (grouped, iterated summation; K = power of 2), with elementary trigonometric periodicity for the half/quarter reductions.
• The central motivation is a recognized, predictable hardware-vs-operation-count trade-off — exactly the "small-sized processing unit" problem the '421 specification itself identifies, and the N/2-complex-work-area drawback Bringham identifies.
• Claims 5 and 12 are the best-defended claims because their exact coefficient expressions are not shown in any one reference; a challenger should attach a step-by-step algebraic derivation from equation (1) to carry §103, otherwise a fact-finder may treat the specific decomposition as non-obvious.
• Flagged uncertainties: (i) the CITATION "JPH0444299A/Yokohama Rubber wave absorber" is a mis-mapping of JP Kokai 4-44099; (ii) US 5,349,549's §103(c) common-ownership treatment for this 1993-chain application is a nuance I cannot confirm to certainty and that could defeat Grounds relying on it; (iii) the "solely real number data" limitation's scope (narrow = disclosed by US 5,349,549; broad = a real-vs-complex distinction weakened by the spec's own simplicity rationale).

Confidence: High on the identity, dates, and technical content of the cited references and on the claim-mapping for claims 1–4, 6–11, 13–14. Medium on claims 5/12 (turn on how "obvious to try"/mechanical algebra is weighed). Lower on the §103(c) common-ownership point and on the exact scope to be given '421's "pre-processing section" — both are explicitly flagged rather than resolved.