US 7979277

I'll research this patent, including any 2026 CAFC docket activity. Let me run several searches.

Let me get the full claim text and check the specific CAFC docket 24-2206 status.

US Patent 7,979,277 — Analyst Summary

Caveat up front: I searched public web sources (Google Patents, PTAB/PTACTS papers, the CAFC website, Law360/Fed Circuit Blog, and district-court filings) rather than issuing live authenticated queries to USPTO PatentCenter and the CAFC docket system. Where I could not confirm something directly, I say so explicitly. I did not return or merge data for similarly numbered patents (e.g., 10,062,377, 10,839,789, 7,587,319, 10,971,140), which belong to the same Zentian family but are different patents.

Bibliographic data (from the patent text provided)

Abstract (verbatim)

"A speech recognition circuit comprising a circuit for providing state identifiers which identify states corresponding to nodes or groups of adjacent nodes in a lexical tree, and for providing scores corresponding to said state identifiers, the lexical tree comprising a model of words; a memory structure for receiving and storing state identifiers identified by a node identifier identifying a node or group of adjacent nodes, said memory structure being adapted to allow lookup to identify particular state identifiers, reading of the scores corresponding to the state identifiers, and writing back of the scores to the memory structure after modification of the scores; an accumulator for receiving score updates corresponding to particular state identifiers from a score update generating circuit which generates the score updates using audio input, for receiving scores from the memory structure, and for modifying said scores by adding said score updates to said scores; and a selector circuit for selecting at least one node or group of adjacent nodes of the lexical tree according to said scores."

Overview of the technology

The disclosure targets embedded/mobile speech recognition (phones, PDAs, dictation devices) where memory, power, and die area are scarce. Its central architectural idea is a CAM/RAM "phone instance network engine" (PINE) in which speech-recognition data structures are mapped onto a content-addressable memory: the CAM stores phone-instance numbers and HMM state IDs (searched by data), and the associated RAM stores token scores (addressed by the CAM output). Tokens are propagated, pruned (beam/word-end/max-model pruning), and passed between phone instances. Supporting blocks include a wave-front manager, phone book (lexical tree + triphone generation), path-merge unit, word engine/word lattice, grammar engine, language-model factor ("smearing"/lookahead) interface, and a distance-calculation accelerator (Mahalanobis distance engine) with compressed acoustic models, a decompressor, CRC/checksum validation, and dual A/B result memories enabling pipelined front-end → distance → search operation.

Independent claims — plain language

The full claim set is not reproduced in the patent text I was given (it ends within the description). The claim language below is taken from PTAB and district-court documents that quote the patent, not from the patent's own claim listing, so wording should be treated as high-confidence but not a substitute for the granted claims.

The '277 patent has 16 claims; independent claims are 1, 14, 15, and 16 (asserted claims in litigation: 1, 4, 5, 7, 9, 10, 12, 14–16).

Claim 1 – speech recognition circuit (multi-processor pipelined version). A circuit with: (a) an audio front end that computes a feature vector from an audio signal (the feature vector being extracted and/or derived quantities over a defined audio time frame); (b) a calculating circuit that computes distances indicating similarity between a feature vector and multiple predetermined acoustic states of an acoustic model; (c) a search stage that uses those distances to identify words in a lexical tree (the tree being a model of words); wherein the front end and search stage run on a first processor and the calculating circuit runs on a second processor; and wherein data is pipelined from front end → calculating circuit → search stage.
(Quoted in Zentian's Patent Owner Response materials for IPR2023-00034; see PTACTS document.)

Claim 14 – speech recognition circuit (connection/pipelining version). A circuit with: (1) an audio front end for computing a feature vector (extracted/derived quantities over an audio time frame); (2) calculating means for computing a distance indicating similarity between a feature vector and a predetermined acoustic state of an acoustic model; (3) a search stage that uses the calculated distances to identify words in a lexical tree (model of words); wherein the audio front end, the calculating means, and the search stage "are connected to each other to enable pipelined data flow." Note this claim uses means-plus-function ("calculating means") and does not require the two-processor split of claim 1.
(Quoted in the Amazon complaint, W.D. Tex. 6:22-cv-00123.)

Claim 15 – speech recognition method. A method counterpart to the circuit claims. I could not retrieve the verbatim granted text of claim 15. The specification's summary of a corresponding "further aspect" describes steps of: storing state identifiers (states corresponding to nodes or groups of adjacent nodes in a lexical tree) and their scores in a memory structure that permits lookup, score reading, and write-back of modified scores; and, repeating for each incoming audio frame: obtaining score updates for the frame, accessing the memory structure to read/update/write back scores by adding the updates, determining whether scores for states furthest on in the lexical tree correspond to a significant likelihood, and if so accessing the lexical tree to determine the next set of possible states. Treat this as an informed paraphrase of the disclosed method aspect, not a quotation of claim 15. Claim 15 was challenged in IPR2023-01193 and was among the asserted claims.

Claim 16 – non-transitory storage medium. A non-transitory storage medium storing processor-implementable code that, when executed, controls at least one processor to implement a speech recognition method (per the Amazon complaint's characterization of claim 16).

Litigation and administrative status (relevant to your CAFC 2026 request)

• IPR2023-01193 — [Amazon.com Services LLC](/litigations/by-plaintiff/Amazon.com%20Services%20LLC) v. Zentian Limited, on US 7,979,277. Filed 2023-07-12; instituted 2024-01-10; Final Written Decision 2024-06-11 (Panel: Ogden, Smith, Turner). Challenged claims 1, 4, 5, 7, 9, 10, 12, 14, 15, 16. (Google Patents lists this proceeding as "PTAB case IPR2023-01193 filed (Final Written Decision).")
• IPR2023-00034 — Apple's parallel IPR against the '277 patent; the disputed issue at oral argument (2024-03-12) was whether the claimed pipelining is compatible with feedback-based pruning from the search stage back to the distance-calculation stage.
• CAFC Appeal 24-2206 — Zentian Ltd. v. Apple Inc., appeal from the PTO. The only CAFC item I could confirm for docket 24-2206 is a nonprecedential order dated 2025-02-14 (cafc.uscourts.gov). I did not find a 2026 merits disposition specifically for 24-2206 in my searches, so I cannot state its current outcome with confidence.
• Other Zentian CAFC activity in 2026 (different patents, listed only to avoid confusion): 24-2204 (Zentian v. Apple, concerning the '319 patent; affirmed under Fed. Cir. R. 36 on 2026-06-08); a 2026-09-25 nonprecedential affirmance in Zentian Ltd. v. Apple finding all claims of US 10,062,377 unpatentable; and Law360 coverage on 2026-07-21 of a Federal Circuit revival of an Apple invalidity challenge to a Zentian speech patent in the California suit. None of these three has been confirmed as directed to the '277 patent.
• District court: Asserted (with other Zentian patents) against Apple in N.D. Cal. (3:23-cv-02921) and against Amazon in W.D. Tex. (6:22-cv-00123, 6:22-cv-00122). Key disputed term for the '277 patent: "pipelined data flow" (claim 14). Zentian is characterized in third-party commentary as an NPE; Unified Patents ran a prior-art crowdsourcing contest on at least claim 15.

Uncertainty notes

1. Claim text sourcing. The authoritative patent text I was given does not include the claims section; claim 1, 14, and 16 wording is quoted from PTAB/district-court papers, and claim 15 is a paraphrase. Verify against the granted claims before relying on exact language.
2. Claim 15. I do not have its verbatim text — flagged rather than reconstructed.
3. CAFC 24-2206 in 2026. I could not confirm a 2026 merits decision for this specific appeal; the only confirmed entry is the 2025-02-14 order. The June/July/September 2026 Federal Circuit items I found concern other Zentian patents.
4. Priority date. Google Patents labels the 2004-09-14 priority date as an assumption; the UK priority application number (GB 0420464.0) comes from litigation filings, and defendants have challenged Zentian's priority entitlement in invalidity contentions.
5. Legal status. "Active, expires 2028-07-23" is Google Patents' assumption, not a legal conclusion.

Primary sources: https://patents.google.com/patent/US7979277/en · https://portal.unifiedpatents.com/ptab/case/IPR2023-01193 · https://www.cafc.uscourts.gov/02-14-2025-24-2206-zentian-ltd-v-apple-inc-order-24-2206-order-2-14-2025_2468192/ · https://cases.justia.com/federal/appellate-courts/cafc/24-2204/24-2204-2026-06-08.pdf · https://fedcircuitblog.com/2026/09/25/opinions-orders-september-25-2026/ · https://patroll.unifiedpatents.com/contests/EAotDyHpR75AHY2wm · https://ai-lab.exparte.com/documents/dct/[4379492](/patent/4379492)/6-22-cv-00123_txwd_complaint_exparte.pdf · https://ai-lab.exparte.com/case/dct/cand/3:23-cv-02921/doc/1 · https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1550429](/patent/1550429)/download-documents

I'll research the litigation history for US 7979277 using the sources you specified.

Let me gather more details on the specific district court cases and the Federal Circuit appeals.

Litigation Involving U.S. Patent No. 7,979,277 ("the '277 patent")

Caution / scope note: I interpreted the patent number literally as 7,979,277 and screened out results for similar numbers. The '277 patent is part of a family that also includes U.S. Patent Nos. 7,587,319, 10,062,377, 10,839,789, and 10,971,140 — those patents appear in the same suits discussed below, so I distinguish where the '277 patent itself was actually asserted. I've separated district court litigation from PTAB proceedings and Federal Circuit appeals, since the latter are administrative/appellate rather than new district-court cases.

A. District Court Litigation (patent infringement)

1. Zentian Ltd. v. Apple Inc.

• Plaintiff: Zentian Ltd. (UK)
• Defendant: Apple Inc.
• Jurisdiction / Court: U.S. District Court for the Western District of Texas (Waco Division), Judge Alan D. Albright
• Case No.: 6:22-cv-00122-ADA
• Filing Date: February 2, 2022
• Status/Outcome: Transferred to the U.S. District Court for the Northern District of California in June 2023 (re-docketed as 3:23-cv-02921-TLT, see case 2 below). The '277 patent is one of the five Asserted Patents (per the N.D. Cal. stay stipulation listing the '319, '277, '377, '789, and '140 patents).
• Source: Zentian's '277 IPR mandatory notices (listing "Zentian Ltd v. Apple Inc., 3:23-cv-02921, (N.D.Cal. June 14, 2023)"); N.D. Cal. ECF No. 116 (case transfer/stay record); Fed. Cir. related-cases statement (24-2206).

2. Zentian Ltd. v. Apple Inc. (transferred case)

• Plaintiff: Zentian Ltd.
• Defendant: Apple Inc.
• Jurisdiction / Court: U.S. District Court for the Northern District of California, Judge Trina L. Thompson
• Case No.: 3:23-cv-02921-TLT
• Filing Date: June 14, 2023 (transferred from W.D. Tex. 6:22-cv-00122)
• Status/Outcome: Stayed. On July 26, 2023 the court granted the parties' stipulation to stay pending Apple's IPRs (including IPR2023-00034 on the '277 patent). After the June 11, 2024 Final Written Decisions, the parties stipulated (June 25, 2024) to maintain the stay to pursue settlement, with a joint status report due September 25, 2024. As of the records available, the case remains stayed subject to settlement/appeal.
• Note: The court considered and weighed Apple's §1404(a) transfer motion and related witness-location evidence; the case was transferred out of W.D. Tex. to N.D. Cal. The complaint accused Apple products (iPhone, iPad, Watch, AirPods, HomePod, Apple TV, Macs) and, as to the '277 patent, Apple's on-device "Hey Siri" wake-phrase and dictation functionality plus the Neural Engine / Always On Processor architecture.
• Sources: N.D. Cal. ECF Nos. 116 (stay stipulation, July 2023) and 125 (stipulation maintaining stay, June 25, 2024); Ex Parte AI case summary for 3:23-cv-02921.

3. Zentian Ltd. v. [Amazon.com, Inc.](/litigations/by-plaintiff/Amazon.com%2C%20Inc.) (later Amazon.com Services LLC and Amazon Web Services, Inc.)

• Plaintiff: Zentian Ltd.
• Defendant(s): Originally Amazon.com, Inc.; by joint stipulation (entered July 21, 2022) substituted with Amazon.com Services LLC and Amazon Web Services, Inc.
• Jurisdiction / Court: U.S. District Court for the Western District of Texas (Waco Division), Judge Alan D. Albright
• Case No.: 6:22-cv-00123-ADA
• Filing Date: February 2, 2022 (original complaint); amended complaint filed August 4, 2022
• Status/Outcome: The '277 patent (claims 1, 4, 5, 7, 9, 10, 12, 14–16) was among the Asserted Patents. A Markman order issued by June 20, 2023 (Dkt. 66). The case was the subject of Amazon's parallel IPR petition (IPR2023-01193). Resolution/current status of the district case beyond the Markman stage is not confirmed in the records retrieved.
• Sources: Amazon's Amended Invalidity Contentions (Ex. 2001 to IPR2023-01197); W.D. Tex. docket 6:22-cv-00123 (Clerk's docket showing the July 21, 2022 substitution order and Aug. 4, 2022 amended complaint); Zentian's '277 IPR mandatory notices.

Possible additional/earlier case (uncorroborated for the '277 patent): Amazon's invalidity contentions incorporate prior art from Zentian Ltd. v. Apple Inc., No. 6:21-cv-815-ADA (W.D. Tex.). That case number predates the two 2022 suits, but I could not confirm from the retrieved records that the '277 patent was asserted in it. Flagging it as a lead rather than a confirmed '277 case.

B. PTAB Inter Partes Reviews on the '277 Patent

These are administrative validity challenges, not district-court suits, but they are the dispositive proceedings driving the district-case stays.

• Both reached Final Written Decisions on June 11, 2024 (Unified Patents PTAB caselist; GreyB/IPverse case records; N.D. Cal. stipulation tables).
• Note: I have not confirmed the specific claim-by-claim patentability outcome for the '277 patent in these two FWDs from the retrieved materials. For context, the companion '319-patent IPR (IPR2023-00033) found all challenged claims unpatentable; the '277 outcomes should be verified directly against the FWDs rather than assumed.
• Sources: https://portal.unifiedpatents.com/ptab/case/IPR2023-01193 ; https://portal.unifiedpatents.com/ptab/case/IPR2023-00034 ; ipverse.greyb.com cases (IPR2023-00034, IPR2023-01193).

C. Federal Circuit Appeals (relating to the '277 patent)

Zentian Ltd. v. Apple Inc., No. 24-2206 (Fed. Cir.)

• Appellant: Zentian Ltd. (Appellees include Apple Inc., Amazon.com Services LLC, Amazon Web Services, Inc.)
• Court: U.S. Court of Appeals for the Federal Circuit
• Case No.: 2024-2206
• Filing Date: August 13, 2024
• Origin: Appeal from PTAB Nos. IPR2023-00034 and IPR2023-01193 (both on the '277 patent)
• Outcome/Status: Dismissed on February 14, 2025, under Fed. R. App. P. 42(b), by agreement of the parties, each side bearing its own costs. (The dismissal by agreement is consistent with settlement activity reflected in the N.D. Cal. stay stipulations.)
• Sources: CAFC Order, 24-2206 (Feb. 14, 2025), https://www.cafc.uscourts.gov/opinions-orders/24-2206.ORDER.2-14-2025_2468192.pdf ; Justia docket 24-2206.

Related but NOT '277 appeals (same family, different patents — noted to avoid conflation):

• 24-2204 — IPR2023-00033 / IPR2023-01192 ('319 patent); affirmed (Fed. Cir. June 8, 2026).
• 24-2205 — '377 patent (IPR2023-00035);
• 24-2207 — Apple v. Zentian ('789 patent, IPR2023-00036).
  These involve different patents in the same family and should not be attributed to the '277 patent.

D. Other activity (not litigation, but on the '277 patent)

• Unified Patents "Patroll" prior-art contest on US-7,979,277-B2 (claim 15), a crowdsourced invalidity search contest. This is not a lawsuit. Source: https://patroll.unifiedpatents.com/contests/EAotDyHpR75AHY2wm.
• Google Patents family litigation entry: a Darts-IP "first worldwide family litigation" reference for family 33306576 (the '277 family), which is a family-level index entry rather than a distinct case.

Summary

Three district-court suits implicated the '277 patent — Zentian v. Apple (6:22-cv-00122 W.D. Tex. → 3:23-cv-02921 N.D. Cal.), Zentian v. Amazon (6:22-cv-00123 W.D. Tex.), and the transferred Apple case as a distinct docket — plus two Apple/Amazon IPRs (IPR2023-00034, IPR2023-01193) that reached Final Written Decisions on June 11, 2024 and drove stays of the district cases. The Federal Circuit appeal on the '277 patent (24-2206) was dismissed by agreement on February 14, 2025. If you need the exact claim-level outcomes of the two '277 Final Written Decisions or the current docket status of the Amazon case, say so and I'll pull those specific documents — I did not confirm those two details from the materials retrieved.

I'll start by verifying the PTAB proceedings referenced in the patent's litigation data and searching for claim-level outcomes.

Key proceedings identified. Let me dig into the claim-level outcomes and the appeal.

Let me verify the Amazon joinder proceeding's outcome and check for any additional IPRs or claim-cancellation certificates.

Bottom line up front

The "PTAB proceedings on file" block from the ODP returns no AIA trial proceedings, but that appears to be an ODP ingestion gap, not a factual absence. The patent record's own litigation metadata (in the authoritative full text) lists two PTAB cases with Final Written Decisions, and web-verified Board documents confirm both. I treat the following two as the proceedings on file, and flag the discrepancy.

Proceedings overview

Two AIA trials on file — IPR2023-00034 (Apple) and IPR2023-01193 (Amazon, joined) — both reaching Final Written Decisions on 2024-06-11 that held all 10 challenged claims unpatentable; no claims sustained, nothing settled, no active proceeding, and the sole appeal (CAFC 2024-2206) was dismissed by agreement on 2025-02-14 without any merits ruling. Net defensive posture: the asserted claims of US 7,979,277 are dead — claims 1, 4, 5, 7, 9, 10, 12, and 14–16 were canceled, which means any demand letter citing those claims (i.e., every asserted claim in the W.D. Tex. and N.D. Cal. complaints) is asserting a patent that no longer has those claims in it. Only a handful of never-challenged dependent claims remain, and I could not verify their current status.

⚠️ Source conflict flagged: The structured ODP block says no AIA proceedings. The patent's own litigation metadata in the full text and independent web verification (PTAB public portal documents, CAFC order PDF, docket aggregators) confirm the two IPRs below. Verify the post-FWD certificates in PTAB E2E before relying on claim status in a filing.

IPR2023-00034 — Apple Inc. v. Zentian Limited

• Type: Inter Partes Review (35 U.S.C. §§ 311–319)
• Filed: 2022-11-15
• Status: Final Written Decision (2011 FWD issued 2024-06-11); appeal dismissed 2025-02-14 → decision final
• Judge panel: Kevin F. Turner, Jeffrey S. Smith, Christopher L. Ogden (Ogden authored)
• Petition grounds: Challenged claims 1, 4, 5, 7, 9, 10, 12, 14, 15, 16 under pre-AIA § 103(a) only. Ground 1: claim 1 (and dependents) over Jiang (US 6,374,219) in view of Baumgartner (US 2002/0049582 A1) and Brown, with or without Smyth (US 5,819,222); claim 4 adds Kazeroonian; claims 9 and 10 add Vensko (US 4,567,606). Eight grounds in total (each combination with and without Smyth). Petitioner expert: Christopher Schmandt (Ex. 1003); Patent Owner expert: David Anderson, Ph.D. (Ex. 2020).
• Institution decision: Instituted on all challenged claims and all grounds — 2023-06-12 (Paper 10). The Board did not deny discretionarily; Apple had filed a Sotera stipulation, which the panel credited in the § 314(a)/Fintiv analysis.
• Final Written Decision: Verdict at claim level — every challenged claim unpatentable. The Board's ordering language: "ORDERED that claims 1, 4, 5, 7, 9, 10, 12, and 14-16 of the '277 patent are unpatentable." Reasoning highlights:
  • Claim 1: "we find Apple's arguments persuasive that the combination of Jiang, Baumgartner, and Brown, with or without Smyth, teach all the limitations of claim 1 and that a person of ordinary skill in the art would have had reason to combine the references as Apple proposes, with a reasonable expectation of success."
  • Zentian's sole substantive theory was that Jiang's pruning is feedback-based and therefore incompatible with Brown's pipelining. The Board rejected that as bodily incorporation of an unclaimed limitation: "pruning is not required as part of the combination" (Jiang teaches pruning as optional, col. 8:52), and the claims do not recite feedback. Zentian's own '319 patent was cited against it as describing search-stage pruning without feedback.
  • No claim was held patentable; the FWD did not reach claims 2, 3, 6, 8, 11, 13 or any claim above 16.
  • Oral hearing: 2024-03-12.
• Settlement / termination: None. Trial ran to FWD.
• Appeal: Zentian Ltd. v. Apple Inc., No. 2024-2206 (Fed. Cir.) — Notice of Appeal filed 2024-08-09; the appeal covered both IPR2023-00034 and IPR2023-01193. By nonprecedential order dated 2025-02-14, the court dismissed under Fed. R. App. P. 42(b) ("The parties having so agreed") with each side bearing its own costs; issued as the mandate the same day. No merits ruling — the Federal Circuit never construed "feature vector" or the pipelining limitations in this appeal. (CAFC order PDF; Justia)
• Defensive value: This is the worst-case outcome for the patent owner and the best for a defendant — every asserted claim is canceled on § 103(a) grounds, and the invalidity went unreviewed because Zentian walked away from its own appeal. Any infringement theory built on claims 1, 4, 5, 7, 9, 10, 12, 14, 15, or 16 is now built on canceled claims.

IPR2023-01193 — [Amazon.com Services LLC](/litigations/by-plaintiff/Amazon.com%20Services%20LLC) (with Amazon Web Services, Inc.) v. Zentian Limited

• Type: Inter Partes Review, instituted and joined to IPR2023-00034 under 35 U.S.C. § 315(c) / 37 C.F.R. § 42.122
• Filed: 2023-07-12
• Status: Final Written Decision (2024-06-11); terminated 2024-06-11
• Judge panel: Christopher L. Ogden, Jeffrey S. Smith, Kevin F. Turner (same panel as the Apple IPR)
• Petition grounds: Same claims 1, 4, 5, 7, 9, 10, 12, 14–16, same grounds — Amazon's petition was, by its own admission, "intentionally identical in substance to the Apple Petition," relying on substantially the same Schmandt testimony.
• Institution decision: Instituted and joinder granted — 2024-01-10. The panel expressly rejected Zentian's argument that Amazon would be time-barred under § 315(b) absent joinder (Zentian was served in the Feb. 2022 W.D. Tex. complaint; Amazon petitioned July 2023), and rejected Zentian's demand that Amazon match Apple's Sotera stipulation: "no such stipulation is required where the Fintiv factors as a whole do not otherwise weigh in favor of discretionary denial." Institution was on the same claims and grounds as the Apple IPR, with the grounds "unchanged and remain[ing] the only grounds on which trial has been instituted."
• Final Written Decision: Same claim-level disposition as IPR2023-00034 — claims 1, 4, 5, 7, 9, 10, 12, and 14–16 unpatentable under § 103(a). It is a combined judgment covering both case numbers.
• Settlement / termination: None; no adverse-judgment or settlement termination. No settlement terms exist to be confidential.
• Appeal: Covered by the same No. 2024-2206 appeal and the same 2025-02-14 dismissal order.
• Defensive value: Confirms the invalidity is not an Apple-specific artifact — a second, independent petitioner (with a different lead counsel team, Fenwick & West) reached the identical result on the identical record. Amazon's joinder also means Amazon is now estopped alongside Apple.

Strategic summary

Claim status. Of the ten claims Zentian actually asserted across Zentian v. Apple (N.D. Cal. No. 3:23-cv-02921, transferred from W.D. Tex. 6:22-cv-00122) and Zentian v. Amazon (W.D. Tex. 6:22-cv-00123), all ten are CANCELED: 1, 4, 5, 7, 9, 10, 12, 14, 15, and 16. Sustained: none — the Board did not uphold a single challenged claim. Untested: claims 2, 3, 6, 8, 11, and 13 (the dependent claims not challenged), plus any claim numbered above 16 if such claims exist (I have not verified the patent's total claim count from the record before me, nor the exact dependency chain of the unchallenged claims, so treat "remaining claims" as unverified). Practically, any unchallenged dependent claim that depends from canceled claim 1 inherits an operative limitation set that no longer exists as an enforceable claim, making it a weak assertion vehicle — but it is not formally canceled, and I could not confirm the certificate of cancellation date under 35 U.S.C. § 318(b).

Estoppel landscape. Estoppel under § 315(e)(2) attached to Apple and Amazon (and their real parties in interest and privies) on the date of the FWD, 2024-06-11, and is now locked in by the 2025-02-14 dismissal with the mandate issued. Those parties cannot raise in district court any § 102/§ 103 ground they raised or reasonably could have raised on claims 1, 4, 5, 7, 9, 10, 12, 14–16. For any other defendant being asserted today: no statutory estoppel runs against you, and the full prior-art space is open — including § 101 eligibility and § 112 written-description/definiteness attacks (which Apple and Amazon had to leave at the district court door and which appear in Defendants' invalidity contentions). More importantly, you are defending claims that no longer exist; the Board's FWD plus the un-appealed finality make a Blonder-Tongue-style collateral attack straightforward if any surviving dependent claim is ever asserted.

Pattern signals. This patent drew exactly two petitions, both from the same litigation cluster (Feb. 2022 W.D. Tex. complaints) — Apple as the lead challenger, Amazon as a § 315(c) joiner after its own § 315(b) bar. Same panel (Ogden/Turner/Smith) decided both. No defensive aggregator filed on the '277: "Unified Patents" appears in the patent record only as the data vendor for the litigation dataset, not as a petitioner. Zentian litigated hard (Noroozi PC as Patent Owner counsel; Bartlit Beck local), pursued its PTAB appeal, then voluntarily dismissed it — consistent with the parallel district cases resolving (the Amazon case terminated 2024-11-20; the Apple N.D. Cal. case had been stayed pending the IPRs). Zentian's aggressive appeal posture continued on sibling patents (the '789 appeal where the CAFC vacated and remanded, and the '377 appeal affirmed 2026-09-25) — but not on the '277, which tells you the '277 claims were conceded as lost.

Recommended next steps

1. If you received a demand or complaint citing US 7,979,277: check the asserted claim numbers first. If they are 1, 4, 5, 7, 9, 10, 12, 14, 15, or 16, the disposition is dispositive — the Board ordered exactly that set unpatentable: "ORDERED that claims 1, 4, 5, 7, 9, 10, 12, and 14-16 of the '277 patent are unpatentable." Read the FWD via the PTAB public portal and the CAFC dismissal order (24-2206 order) and put the cancellation status in your Rule 11 / § 285 correspondence. Case-level entries: IPR2023-00034, IPR2023-01193.
2. Verify the paper trail before filing anything: pull PTAB E2E for both case numbers to obtain (a) the FWD PDF, (b) confirmation that no rehearing request remained pending, and (c) the § 318(b) certificate of cancellation for the '277 claims. My sources confirm the FWD and the appeal dismissal, but I could not independently confirm the certificate issuance date — treat that as an open item.
3. No active proceedings to monitor. Both trials terminated 2024-06-11; the appeal terminated 2025-02-14; the W.D. Tex. Amazon action terminated 2024-11-20. There is no institution deadline, oral hearing, or statutory § 316(a)(11) FWD date outstanding on this patent.
4. If a surviving dependent claim (2, 3, 6, 8, 11, or 13) is ever asserted: demand proof of the claim's dependency chain and current status, and consider a collateral-estoppel motion grounded in the FWD. Because the FWD did not address those claims, do not represent that the Board invalidated them — it did not.
5. If you are Apple or Amazon (or in privity): § 315(e)(2) estoppel now bars the § 102/§ 103 grounds you raised or reasonably could have raised; confine any further challenge to § 101/§ 112 or to non-patent/non-printed-publication art.

Confidence / caveats. High confidence on proceeding numbers, petitioner identities, panels, dates, institution posture, and the claim-level FWD disposition (all corroborated by Board documents and the CAFC order). Medium confidence on the exact ground-to-claim mapping (reconstructed from the FWD text as surfaced, and the ground list is partially inferred); low/unverified on the certificate of cancellation, the '277 patent's total claim count, and the dependency chain of the unchallenged claims. The ODP "no proceedings" result is contradicted by the authoritative patent text and by public PTAB/CAFC records — treat it as an ingestion gap.

I'll research this patent's ownership chain, litigation, and any NPE signals before drafting.

Let me dig into the assignment records and the NPE directories.

Assignment & Ownership Analysis — US 7,979,277 B2

Scope note / data limitation: Google Patents' legal-events record for this patent shows exactly one recorded ownership transfer — a 2007-03-14 reassignment from the four inventors to Zentian Limited. Within this session I was unable to retrieve the reel/frame number or correspondent of record for that recording from the USPTO Assignment Center (assignmentcenter.uspto.gov / assignment.uspto.gov). The findings below are grounded in the Google Patents legal-events entry plus litigation/PTAB records; where I could not verify a reel/frame, I say so rather than invent one.

Inventors

Unusual pattern: no inventor attrition signal. All four were Zentian personnel at filing, and the patent's inventor-of-record set stayed intact across the entire family (US 8,352,262; 9,076,441; 10,062,377; 10,839,789 all name the same four). Zentian's own complaint (W.D. Tex. 6:22-cv-00122, ¶34) states most work was unpaid — done for share options, because funding was hard to raise post-dot-com. That is a "bootstrapped startup" pattern, not the "inventors depart within 12 months → portfolio fire-sale" pattern.

Original assignee

Zentian Limited (Zentian Ltd), a private company incorporated in the UK. Registered address per complaint ¶16: 27 Queens Close, St. Ives, Cambridgeshire, England, PE37 5QD — a residential-style address, not a corporate campus or registered-agent suite.

• Primary line of business: R&D of a hardware speech-recognition SoC / acoustic accelerator (FPGA prototype paired with CMU Sphinx, ~2× speedup per complaint ¶36).
• Did it ship a product embodying the claims? No. It marketed the accelerator to Nuance, IBM, Sensory, Samsung, Qualcomm and carriers and was repeatedly declined (complaint ¶¶38–39). The W.D. Tex. venue-transfer order (July 2023) states: "Zentian is not engaged in product competition in the marketplace." Only a partial prototype was ever built.
• Current status: Operating as a licensing/enforcement entity. It remains the named owner of the patents and is litigating; there is no evidence it ever dissolved or filed bankruptcy. Unified Patents' PTAB case list classifies it as "NPE (Small Company)."

Assignment timeline

• 2007-03-14 (recorded; execution date not surfaced separately) / recorded 2007-03-14 — Reel/Frame not retrievable in this session
  • Conveyance: Assignment ("Assignment of Assignors' Interest")
  • Assignor: Catchpole, Mark; Harris-Dowsett, Damian K.; Larri, Guy; Reynolds, Timothy B. (the four inventors)
  • Assignee: Zentian Limited
  • Correspondent: Not obtained — the USPTO Assignment Center record could not be pulled within this session, so I cannot name the recording attorney/firm. Conversely, I can confirm no recurring assignment correspondent appears, because there is only one recorded assignment in the chain.
  • Context: Founder/inventor-to-company assignment — the standard transfer of inventor rights to the startup that employed them; timing coincides with US national-stage entry of PCT/GB2005/003554 (filed 2005-09-14; application 11/662,704 filed 2007-03-14).

No post-issuance assignment is recorded. Google Patents' legal events show no transfer to any LLC, aggregator, or successor after issuance (2011-07-12). This is corroborated by Zentian's own pleadings: complaint ¶17 in both Zentian v. Apple (6:22-cv-00122) and Zentian v. Amazon (6:22-cv-00123) states "Zentian is the owner of all rights, title, and interest in the Asserted Patents." So the chain is a single link: inventors → Zentian Limited, unchanged for ~19 years.

Timeline diagram

timeline
    title Ownership of US 7979277
    2004 : Priority date established
    2005 : PCT application filed
    2007 : US national stage filed
         : Inventors assign rights to Zentian Ltd
    2011 : US 7979277 issued
    2022 : Zentian sues Apple and Amazon
    2023 : Apple and Amazon file IPRs
    2024 : PTAB final written decisions
    2025 : Federal Circuit appeal dismissed

NPE / troll-pattern signals

1. Shell-entity transfer — not present. There is no transfer from an operating assignee to a licensing-only LLC. The single recorded assignment (2007-03-14) runs inventors → Zentian Limited, and Zentian is the original assignee, not a downstream shell. The "no products in commerce" element is satisfied (court finding, July 2023; no commercialization per complaint ¶38–39), but the transfer element is absent, so the signal as defined does not fire.

2. Known asserter in the chain — present (as a small-company NPE). Zentian Limited is not on the enumerated high-frequency lists (Acacia, Marathon, IV, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp, Spangenberg). However, Unified Patents expressly designates it an NPE: its PATROLL post of 2022-05-27 reads "U.S. Patent 7,979,277, owned by Zentian Limited, an NPE… asserted against Apple and Amazon," and Unified's PTAB portal tags the owner "NPE (Small Company)." Assertions: Zentian Ltd v. Apple (6:22-cv-00122) and Zentian Ltd v. Amazon (6:22-cv-00123), both filed 2022-02-02 in W.D. Tex.

3. Repeat correspondent across the chain — unclear / insufficient data. With only one recorded assignment, there is no chain over which a correspondent could recur, and I could not retrieve even that single correspondent. Flag but do not count: litigation/PTAB counsel Noroozi PC (Peter Knops) appears as Zentian's patent-owner counsel across the IPR2023-00033/00034/00035/00036 and IPR2023-01192–01197 family. That is litigation counsel, not an assignment correspondent, and is outside the signal as defined.

4. Cascading transfers — not present. No chained LLC assignments within 24 months or any other window. One assignment in ~19 years.

5. Pre-litigation transfer — not present. The suits were filed 2022-02-02; the only assignment is dated 2007-03-14 — roughly 15 years earlier. Nothing was arranged on the eve of litigation.

6. Bankruptcy fire-sale — not present (no evidence). Zentian's pleadings describe chronic underfunding and unpaid engineering (complaint ¶34), but there is no evidence of a Chapter 7/11 filing or a court-supervised patent sale. Treat as not present on the available record.

7. Privateering — not present. Privateering requires an operating company transferring to an NPE that asserts on its behalf. Here the NPE is the original R&D company asserting its own patents; no operating-company patron is in the chain.

8. Defensive aggregator (anti-NPE) — not present. The chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. The opposite: the patent was asserted against Apple and Amazon, drawing Apple-initiated IPRs (IPR2023-00034, petition 2022-11-15) and Amazon-initiated IPRs (IPR2023-01193, filed 2023-07-12, joined to Apple's), which reached final written decisions on 2024-06-11 and a Federal Circuit appeal (No. 24-2206) dismissed on 2025-02-14.

Verdict

NPE — moderate confidence.

The current and original assignee is Zentian Limited, which the W.D. Tex. transfer order (July 2023) found is "not engaged in product competition in the marketplace," and which asserted the '277 patent against Apple and Amazon on 2022-02-02 without ever having shipped a product embodying the claims (only an FPGA prototype; complaint ¶¶36–39). That is one strong signal — signal 2, because Unified Patents designates Zentian an NPE in its 2022-05-27 PATROLL contest and its PTAB portal ("NPE (Small Company)"). The transfer-based signals (1, 4, 5, 6, 7, 8) do not fire, because the chain is a single 2007-03-14 inventor→company assignment that never moved — this is a failed startup monetizing its own patents, not the classic shell/roll-up pattern, which is why I stop short of high confidence. Because only one signal clearly fires, moderate confidence is the defensible call under the rubric.

Verify at the USPTO Assignment Center: https://assignment.uspto.gov/patent/index.html (search patent 7,979,277), and cross-check Google Patents legal events: https://patents.google.com/patent/US7979277/en

Caveat: the single recorded assignment's reel/frame and correspondent of record remain unverified here; retrieving that record is the one concrete gap to close before treating this ownership chain as final.

I'll research US patent 7979277 and its cited prior art. Let me start with searches.

Let me dig deeper into the specific patent citations and the IPR references.

I have substantial information. Let me now identify the key primary references (Jiang, Brown) and the front-page citations.

Prior-Art Analysis — US Patent No. 7,979,277 B2

1. Patent identification (as literally recorded)

Source: https://patents.google.com/patent/US7979277/en

2. Methodology and an important limitation (stated explicitly)

I searched for the patent-face "References Cited" (U.S./foreign patent documents and NPL) printed on the '277 patent. My searches did not return a verifiable copy of that front-page citation list for the '277 patent itself. Several results returned citation lists belonging to other patents (e.g., US 7,587,318 to Seshadri) and to unrelated WO publications — those are not the '277 citations and I have not attributed them to '277.

What I could verify, from USPTO PTAB petition records and litigation filings, is the prior art actually applied against the '277 patent in the PTAB and district court. That is the substantively important body of prior art, and it is what I report below. Where I state a citation, I am grounding it in a retrieved record; where I could not confirm a number, I say so rather than inventing one.

Caveat on §102 vs. §103: Almost all of the '277 prior art was applied in obviousness (§103) combinations, not as standalone §102 anticipation. I flag each reference accordingly — a reference used only in a combination does not by itself "anticipate" a claim under §102.

3. The controlling PTAB/litigation prior art

3.1 Primary reference — "Jiang"

• Citation: U.S. patent to Jiang (specific patent number not confirmed in the records I retrieved; it is the primary reference in Apple's IPR2023-00034 grounds).
• Disclosure: Large-vocabulary speech recognition system having a feature extraction module 66, a lexicon 70, and a tree search engine 74 performing a lexical-tree search; uses codewords with "discrete and semi-continuous hidden Markov modeling," and its specification contemplates direct use of feature vectors in distance calculations (Ex. 1004, 1:21–25, 2:5–6, 7:30–42, 7:30–8:43).
• Mapped by petitioner to: claim 1(a)–(d) — audio front end / feature vector; calculating circuit producing distances; and search stage using calculated distances to identify words within a lexical tree (37:… 38:20–22).
• Status: The Board found Apple's showing persuasive; combined with Baumgartner and Brown, it renders claim 1 (and, with added references, claims 4, 5, 9, 10) unpatentable under §103(a). Jiang is a §103 primary reference, not a standalone §102 anticipator.

3.2 U.S. Patent No. 5,819,222 — Smyth et al. ("Smyth")

• Citation: U.S. Patent No. 5,819,222, issued October 6, 1998.
• Disclosure: Task-constrained speech recognition using direct comparison of feature vectors with continuous-density HMMs. Cited by Apple for the distance-calculation / feature-vector-comparison teaching and used to bolster the Jiang combination.
• Potentially anticipates: none alone. Used as a §103 secondary reference (Zentian argued the Smyth+Jiang combination was improper because of the continuous-density vs. discrete/semi-continuous HMM mismatch; the Board nonetheless found the combination persuasive).

3.3 U.S. Pub. No. 2002/0049582 — Baumgartner et al. ("Baumgartner")

• Citation: U.S. Patent Application Publication No. 2002/0049582 A1.
• Disclosure: Teaches a front end and search stage provided on a first processor and a distance-calculating circuit on a second processor.
• Mapped by petitioner to: claim 1(e) — "wherein said audio front end and said search stage are implemented using a first processor, and said calculating circuit is implemented using a second processor" (38:23–25).
• Potentially anticipates: none alone — §103 secondary reference for the split-processor architecture.

3.4 "Brown"

• Citation: U.S. patent/publication to Brown (specific number not confirmed in the retrieved records).
• Disclosure: Pipelining of recognition stages; cited for concurrent pipelined operation of the distance-calculation and search stages.
• Mapped to: claim 1 (pipelining) and claim 4 (multi-threading), in combination with Jiang and Baumgartner.
• Potentially anticipates: none alone — §103.

3.5 U.S. Patent No. 4,567,606 — Vensko ("Vensko")

• Citation: U.S. Patent No. 4,567,606, issued January 28, 1986. (Ex. 1038)
• Disclosure: "A continuous speech recognition system [that] includes a plurality of processors doing template comparisons of speech data" (57). Generates interrupts to inform processors that results are ready for transfer (6:22–29, 6:48–51, 10:10–19).
• Mapped by petitioner to: claims 9 and 10 —
  • Claim 9: "wherein the speech accelerator has an interrupt signal to inform the front end that the accelerator is ready to receive a next feature vector from the front end" (38:54–57).
  • Claim 10: "wherein the accelerator signals to the search stage when the distances for a new frame are available in a result memory" (38:58–60).
• Potentially anticipates: none alone — §103 secondary reference.

3.6 "Kazeroonian"

• Citation: U.S. patent to Kazeroonian (specific number not confirmed; Ex. 1037).
• Disclosure: Multi-threaded operation dividing tasks among separate processing threads (15:33–16:1, 19:26–20:5, 21:9–11).
• Mapped to: claim 4 (separate threads), combined with Jiang+Baumgartner+Brown.
• Potentially anticipates: none alone — §103.

3.7 U.S. Patent No. 6,959,376 B1 — Boike et al. ("Boike")

• Citation: U.S. Patent No. 6,959,376 B1, issued October 25, 2005. (Ex. 1006)
• Disclosure: Speech-recognition/processing architecture cited by petitioner among the asserted prior-art set.
• Potentially anticipates: cited in the IPR ground set; no single-reference anticipation established in the record I retrieved.

3.8 U.S. Patent No. 4,926,488 — Nadas et al. ("Nadas")

• Citation: U.S. Patent No. 4,926,488. (Ex. 1019)
• Disclosure: Relates to prototype vectors (characterized in the litigation as codewords) and distance-based comparison of feature vectors to prototype vectors — "the feature vector for each time interval is typically compared to each prototype vector based on a predetermined closeness measure… the closest prototype vector is selected" (col. 1).
• Relevance: central to the claim-construction fight over whether "feature vector" includes codewords (see §4 below). Used as a §103/NPL support reference, not a standalone anticipator.

3.9 U.S. Patent No. 6,182,037 — Maes ("Maes")

• Citation: U.S. Patent No. 6,182,037. (Ex. 1023)
• Disclosure: Multi-modal (audio+video) speech processing; cited in related Zentian IPRs (e.g., the '330/'039 family) for audio/video fusion and alignment, not as a core '277 anticipation reference.

3.10 Additional references carried in the '277 IPR exhibit set

• U.S. Patent No. 5,774,859 — Houser (issued 1998-06-30; Ex. 1040)
• U.S. Patent No. 7,171,358 B2 — Whittaker (issued 2007-01-30; Ex. 1041)
• U.S. Patent No. 7,478,041 B2 — Ichikawa (issued 2009-01-13; Ex. 1042)
• U.S. Patent No. 7,587,318 B2 — Seshadri (issued 2009-09-08; Ex. 1043)
• U.S. Pub. No. 2002/0091826 A1 — Comeau (published 2002-07-11; Ex. 1044)
• Kevin L. Kloker, "The Motorola DSP56000 Digital Signal Processor," IEEE Micro, Vol. 6, Issue 6, pp. 29–48, December 1986 (Ex. 1009)
• Christopher Schmandt, "Voice Communication with Computers," Van Nostrand Reinhold, 1994 (Ex. 1010)
• Lawrence Rabiner, "Fundamentals of Speech Recognition," Pearson Education, 1993 (Ex. 1015)
• Lawrence R. Rabiner, "A Tutorial on Hidden Markov Models…," Proc. IEEE, Vol. 77, No. 2, pp. 257–286, Feb. 1989 (Ex. 1022)
• John Holmes, "Speech Synthesis and Recognition," 2nd Ed., Taylor & Francis, 2001 (Ex. 1018)
• Frederick Jelinek, "Statistical Methods for Speech Recognition," MIT Press, 1997 (Ex. 1020)
• Shigeru Katagiri, "Handbook of Neural Networks for Speech Processing," Artech House, 2000 (Ex. 1021)
• Richard L. Klevans, "Voice Recognition," Artech House (Ex. 1016)
• Harry Newton, "Newton's Telecom Dictionary," 19th Ed., CMP Books, March 2003 (Ex. 1008)
• Ravishankar, "Efficient Algorithms for Speech Recognition" (Ex. 1012)

These are primarily background/§103-support references (textbooks and handbooks establishing the level of ordinary skill and the state of the art), not standalone §102 anticipators.

4. Which claims each reference potentially reaches (summary table)

No single reference in the retrieved record is a pure §102 anticipation of any independent claim; the invalidity case is built on §103 obviousness combinations.

5. Outcome (as reflected in the retrieved records)

• PTAB, IPR2023-00034 (Apple Inc. v. Zentian Ltd.; also IPR2023-00036, -01193 with Unified Patents): Final Written Decision (dated 2024-06-11/12 in the retrieved docket snippets) holding the challenged claims of the '277 patent unpatentable as obvious over combinations including Jiang, Baumgartner, Brown, Smyth, Vensko, and Kazeroonian. E.g., claim 1 was found obvious over Jiang + Baumgartner + Brown, with or without Smyth; claim 4 over Jiang + Baumgartner + Brown + Kazeroonian; claims 9/10 over Jiang + Baumgartner + Brown + Vensko (optionally + Smyth).
• Federal Circuit: a nonprecedential decision reported as affirming the Board, with the central dispute being the construction of "feature vector" (whether it includes codewords). Docket: Court of Appeals for the Federal Circuit, case 24-2206.
• Litigation: Zentian asserted the '277 patent against Apple and Amazon in W.D. Tex. (6:22-cv-00122 and 6:22-cv-00123) and a N.D. Cal. case (3:23-cv-02921).

6. What I could not confirm — and would verify next

1. The complete printed "References Cited" list on the face of US 7,979,277 — I could not retrieve it; I will not reconstruct it from memory.
2. Exact patent numbers for "Jiang," "Brown," and "Kazeroonian." These were identified in the records only by short name. The Jiang reference is the linchpin of the invalidity case, so its full citation should be pulled directly from Apple's IPR2023-00034 petition (Ex. list at https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1550429](/patent/1550429)/).
3. The Apex Standards prior-art/claim charts referenced in the Unified Patents Patroll contest for claim 15 (https://patroll.unifiedpatents.com/contests/EAotDyHpR75AHY2wm) — these were curated prior art (U.S., non-U.S., and NPL) specifically charted against the '277 memory-structure/CAM-accumulator claim 15, which is the claim most resistant to the Jiang-based grounds.

Note on dates: the retrieved Federal Circuit news item is timestamped Sep 28, 2026, which is consistent with the latest operative date available to me; the "April 26, 2026" in the task header would predate that decision. I am reporting the decision as retrieved rather than reconciling the two dates.

I'll gather grounding on the prior art associated with this patent — including the PTAB proceedings noted on the record — before constructing the § 103 analysis.

Obviousness Analysis — U.S. Patent No. 7,979,277 under 35 U.S.C. § 103

Scope note / caveat up front. The Google Patents record for US 7,979,277 (https://patents.google.com/patent/US7979277/en) carries only a limited "prior art" indication on the page itself: the auto-generated concepts list is essentially the background discussion in the specification, which names Sukun Kim et al. (special-purpose processing elements, one per node or one per state) and Ney et al. (language-model lookahead / "smearing"). Neither Sukun Kim nor Ney was the art that was actually applied against the claims. The operative § 103 art is the art of record in IPR2023‑00034 (Apple v. Zentian) and its joinder IPR2023‑01193 (Amazon v. Zentian). I have grounded the analysis on that record and flag where I could not independently verify a specific mapping. Also note a date discrepancy: your prompt says April 26, 2026, but the system date is 2026‑09‑29, and the search record contains Federal Circuit activity dated September 25–29, 2026; I treat the later date as current.

1. The claims to be assessed

Challenged/asserted claims: 1, 4, 5, 7, 9, 10, 12, 14–16. Claim 1 is representative (text as reproduced in the petition's claim-listing appendix):

Dependents: 4 (first processor multi-threaded; search stage and front end as separate threads); 5 (calculating circuit autonomously calculates distances for every acoustic state); 7 (feature vector = spectral components for a predetermined frame); 9 (accelerator interrupt to tell the front end it is ready for the next feature vector); 10 (accelerator signals the search stage when a new frame's distances are in a result memory); 12 (front end inputs a digital audio signal). Claim 14 is a separate independent claim in means-plus-function form ("calculating means…", "wherein said audio front end, said calculating means, and said search stage are connected to each other to enable pipelined data flow"), i.e. claim 1 without the two-processor requirement.

Critically, the heavy disclosure of the specification — CAM/RAM mapping of the lexical tree, wave-front manager, path-merge unit, phone book, LM-factor engine — is not claimed. The claims are directed at a much broader, three-stage pipelined architecture. That mismatch is the single most important fact in the obviousness analysis.

2. Level of ordinary skill

On the record, the Board treated the level of ordinary skill as including the education and experience necessary to apply speech-recognition techniques in a hardware/system context (the Board expressly found the level of skill "would have included education and experience necessary to apply sp[eech]…" techniques in the Apple proceedings, and Apple's POSITA definition was tied to the field of speech recognition). A reasonable formulation: a bachelor's degree in electrical engineering or computer science (or equivalent) plus several years of experience in speech recognition and/or DSP/hardware architecture. This matters because the asserted claims are, at bottom, allocation of a known three-stage recognition pipeline across processors.

3. The prior art of record and what each reference supplies

The following were also relied on in the related family IPRs (e.g. the '789 patent line), not necessarily against the '277 claims: Boike (US 6,959,376), Houser (US 5,774,859), Whittaker (US 7,171,358), Ichikawa (US 7,478,041), Seshadri (US 7,587,318), Comeau (US 2002/0091826), Mozer, and the non-patent literature Krishna, Ravishankar, Mathew, and Zentian's own '319 patent (Ex. 1067).

Diligence flag: several of those references (Boike, Whittaker, Ichikawa, Seshadri) have issue dates after the '277 priority date of 2004‑09‑14. They can only be prior art under pre‑AIA § 102(e) if their effective filing dates predate the critical date. Any § 103 ground built on them requires that predicate to be established, not assumed.

4. The grounds as actually instituted and adjudicated (IPR2023‑00034)

Per the Board's designation of issues for oral argument and the parallel joinder petition:

1. Claims 1, 5, 7, 12, 14–16 obvious over Jiang + Baumgartner + Brown.
2. Claim 4 obvious over Jiang + Baumgartner + Brown + Kazeroonian.
3. Claims 9 and 10 obvious over Jiang + Baumgartner + Brown + Vensko.
4. Claims 1, 5, 7, 12, 14–16 obvious over Jiang + Baumgartner + Brown + Smyth.
5. Claim 4 obvious over Jiang + Baumgartner + Brown + Smyth + Kazeroonian.
6. Claims 9 and 10 obvious over Jiang + Baumgartner + Brown + Smyth + Vensko.

All challenged claims were held unpatentable in the Final Written Decision of June 11, 2024; the Federal Circuit appeal (No. 24‑2206, covering both IPR2023‑00034 and IPR2023‑01193) was dismissed by agreement on February 14, 2025 (Fed. R. App. P. 42(b)), so there is no merits appellate ruling on the '277 claims. The 2026 Federal Circuit decisions in the Zentian v. Apple nonprecedential line concern the '377 and '789 patents, not the '277.

5. Element-by-element: why claim 1 is obvious

• 1(a)–(b) feature vector from an audio signal. Jiang's front end extracts/vector-quantizes a feature representation per frame. The Board construed "feature vector" broadly — in the related proceedings, to include codewords — and that construction was affirmed on appeal (Sept. 2026). Under either construction, Jiang's per-frame feature representation reads on (a)–(b).
• 1(c) calculating circuit for distances to acoustic states. Jiang's calculation stage computes likelihoods/scores against acoustic models; alternatively, Smyth expressly discloses classifier processor 341 computing state probabilities "indicating the similarity that a feature vector corresponds to each acoustic state in state memory 342." A POSITA would treat a processor performing that function as the claimed "calculating circuit" — the specification assigns no special structure to "circuit."
• 1(d) search stage over a lexical tree. Jiang's tree search engine 74 with memory 72.
• 1(e) first processor (front end + search) / second processor (calculating). Partitioning a recognition pipeline across processors is the core teaching of Baumgartner; and the motivation is self-evident from the art's own framing — distance calculation is the computationally intensive stage, exactly the kind of workload that dedicated processors/DSPs (Kloker) off-load in embedded systems. The Board also relied on Kazeroonian for multi-threading in claim 4, which is the same genus of teaching.
• 1(f) pipelining front end → calculating → search. Pipelining is a classical throughput technique, and the patent specification itself describes the three-stage structure with no more structural detail than the board-level arrangement. KSR permits reliance on "common sense" and on the predictable benefits of known engineering techniques where the result is a mere rearrangement of known elements.

Rationale for combination (KSR-aligned):

1. Known problem / known solution: the distance-computation stage dominates runtime; off-loading it to a second processor is an established remedy (Smyth, Kloker, and the broader embedded-DSP art).
2. Predictable result: pipelining known sequential stages to raise throughput is a design choice with predictable benefits (the specification itself notes 3× throughput from pipelining).
3. Imperfect human recall not required: the references need only "suggest" the arrangement; a POSITA "is also a person of ordinary creativity, not an automaton" (KSR, 550 U.S. at 421), as the Board and Apple's reply brief emphasized.

6. The dependent claims

• Claim 5 (autonomously compute distances for every state): this is the natural consequence of the accelerator reading sequentially through the acoustic model — the specification itself touts exactly this autonomy. Jiang plus a dedicated calculating processor renders it obvious.
• Claim 7 (spectral components of a frame) and Claim 12 (digital audio input): conventional feature-extraction and front-end details; Jiang.
• Claim 4 (multi-threaded first processor; front end and search as separate threads): Kazeroonian teaches dividing tasks among separate threads for performance; Apple argued, and the Board agreed, that running the front end and search as separate threads to avoid data loss was an obvious application of Kazeroonian's teaching.
• Claims 9–10 (interrupt handshakes between stages): Vensko teaches a multi-processor speech recognizer using interrupts to signal that processing results are ready for transfer; applying that to the accelerator/front-end and accelerator/search interfaces is a routine design choice once the stages are separate processors or threads with finite buffers.
• Claim 14 (means-plus-function, pipelined connection without the two-processor limitation): a fortiori obvious; the two-processor limitation is the narrower variant of claim 1, so the Jiang‑Baumgartner‑Brown mapping that defeats claim 1 defeats claim 14 as well.
• Claims 15–16 — I could not retrieve the full text from the sources available to me; the Board did hold them unpatentable within the same grounds. Note that claim 15 attracted a Unified Patents PATROLL bounty contest in May 2022, indicating the industry viewed it as vulnerable.

7. Zentian's counterarguments and why they were unavailing

• Impermissible "bodily incorporation." Zentian's principal attack was that the combination had to import Jiang's feedback-based pruning, which it said was incompatible with pipelining. The Board held that Jiang's pruning is optional ("In addition, a pruning technique can be used," col. 8:52–64), that the claims do not recite pruning, and that the mapping did not require it — so the argument failed as bodily incorporation. Support came from Zentian's own '319 patent (Ex. 1067), which discloses search-stage pruning without feeding results back to the distance stage — an admission that search-stage pruning has independent benefit.
• No reasonable expectation of success / real-time performance / memory collisions. Zentian's expert (Dr. Anderson) offered opinions the Board effectively discounted as conclusory and unsupported by prior art, and the Board credited the references as presumed enabling.
• "Feature vector" construction (codewords). This was the live dispute that carried into 2026. The Board construed "feature vector" to encompass codewords/representative vectors derived from the audio, and the Federal Circuit affirmed that construction in the September 2026 nonprecedential decisions in the sibling cases.

8. What the page's own prior-art statements contribute

The specification's background — Sukun Kim et al. (one processing element per network node or per state) and Ney et al. (LM lookahead/"smearing," reducing active states from ~168,000 to ~8,000) — supports, rather than defeats, obviousness on the claims as issued. Sukun Kim establishes that massively-parallel, special-purpose hardware for the search network was known; Ney establishes that shrinking the active state space to cut computation was known. But standing alone that pair does not disclose the claimed three-way split (front end / separate calculating circuit / search) plus pipelining, so I would not advance Sukun Kim + Ney as the primary ground. Their proper role is (i) evidence of the general knowledge that speech recognition could be distributed across special-purpose processors, supporting a motivation to combine, and (ii) a reason the specification's unclaimed CAM/wave-front detail adds no patentable weight.

9. Bottom line

On the record, the claims of US 7,979,277 (1, 4, 5, 7, 9, 10, 12, 14–16) are obvious under § 103:

• Jiang supplies the three-stage recognition architecture, the lexical-tree search, and the acoustic-model distance computation.
• Baumgartner and Brown supply the multi-processor / pipelined organization that meets 1(e)–(f).
• Smyth supplies an express processor-plus-state-memory "calculating circuit" alternative.
• Kazeroonian supplies multi-threaded execution (claim 4).
• Vensko supplies interrupt handshakes (claims 9–10).

The motivations are the classic KSR ones: a recognized bottleneck (distance computation) with a known remedy (dedicated second processor), and a known throughput technique (pipelining) with predictable results. The Board so held on June 11, 2024, and the appeal was dismissed by agreement, leaving that determination undisturbed.

Confidence and gaps. I am highly confident about the claim text of claims 1, 4, 5, 7, 9, 10 and 14; the identity of the six grounds; the Board's outcome; and the appeal disposition. I have not independently verified the specific Baumgartner and Brown element mappings, the full text of claims 15–16, the full final written decision reasoning on claim 12, or the exact publication dates of Baumgartner and Brown; those should be confirmed against the petition (PTAB Petitions 1550429/1554362) and the June 11, 2024 Final Written Decision before this analysis is relied upon formally.