US 12548565

I'll search for authoritative information on this specific patent number, including any litigation.

Let me verify the specific patent's litigation footprint and family relationships, without conflating similar numbers.

Analyst Report — US 12,548,565 B2

Scope note up front: I searched specifically for the literal string "12548565" / "12,548,565" plus assignee and litigation terms. Searches for the exact number returned no litigation or docket hits. All litigation results that came back relate to other, different patent numbers in the same specification family (see §5). I have not been able to retrieve the granted claim text of US 12,548,565 — the authoritative document I was given is truncated before the claims. Everything in §4 is therefore an overview of the invention's claim categories drawn from the patent's own summary-of-the-invention language, not verified claim language. This is flagged again in §6.

---

1. Bibliographic data

Chain-of-title events of record: security agreement to Wells Fargo Bank, N.A. as collateral agent (recorded 2024-04-15), then a release of that security interest (recorded 2025-01-02).

---

2. Abstract (as published)

The patent is directed to monitoring the acoustic environment of a mobile device. Acoustic input is received from the environment while the device operates in a low-power mode; the device detects whether the acoustic input includes a voice command by performing a plurality of processing stages on the acoustic input, at least one of which is performed while the device is in the low-power mode, and by using at least one contextual cue to assist in that detection. Companion aspects cover a mobile device configured to do this and a computer-readable medium encoding instructions to do this.

---

3. Plain-language technical gist

The problem: conventional voice assistants need a manual trigger (button/icon) and/or an explicit spoken wake word, and are unresponsive while the phone is asleep. Requiring wake-ups and triggers is inconvenient and sometimes impossible (e.g., driving).

The disclosed solution is an "always-on"–style acoustic monitor built around power-tiered, staged decision-making:

• Hierarchical processing — cheap stages (e.g., voice-activity detection: amplitude, duration, energy, spectral slope, periodicity, cepstrum, phoneme/phone-loop matching) run first; expensive stages (limited-vocabulary ASR for triggers/keywords, large-vocabulary ASR, statistical classification, natural-language processing, speaker verification) run only if the cheaper stages cannot rule the input out as "spurious acoustic activity."
• Two-processor split — early stages run on a low-power secondary processor (DSP/co-processor), and the power-hungry primary CPU is engaged only when needed, and only incrementally ("staged wake-up" of display, microphones, GPS, accelerometer, radio).
• Contextual cues — information external to the acoustics (time of day, location, motion/accelerometer, recent device activity/events, battery state, ambient acoustic profile, voice-print/speaker ID, direction-of-arrival via TDOA) is used to bias thresholds, select or skip stages, act as a stage itself, or terminate processing.
• Network offload — some stages may be pushed to servers, if the radio power cost is justified by the savings.
• Selective response — some commands (e.g., music transport) can be executed without leaving low-power mode.

---

4. Overview of the independent-claim categories

Caution: the granted claims were not available to me. The patent's own summary enumerates six independent claim families. I list them below as claim families A–F, based on the summary text, without asserting which claim numbers they occupy or whether all survived into the granted patent.

Key claim-vocabulary definitions that the document supplies and that would control claim scope:

• "explicit trigger" = specific, designated, predetermined actions required to engage the system, encompassing both manual triggers (hands-on-device actions) and explicit voice triggers (a designated wake word/phrase).
• "voice command" = any actionable voice input, including voice requests, voice queries, and explicit voice triggers.
• "low power mode" = any mode entered to conserve power that conventionally requires one or more manual wake-up actions to exit.
• "acoustic input" = information received from the acoustic environment or information derived therefrom (i.e., signals at any stage of analog/digital transformation).
• "spurious acoustic activity" = acoustic activity (including speech) that is not a voice command, including, in some embodiments, speech by people other than the device's user.

---

5. Litigation and 2026 CAFC docket check

Direct result for US 12,548,565: none found. No district-court, ITC, PTAB, or Federal Circuit activity naming the literal number 12,548,565 appeared in my searches. The Google Patents family page does carry a "family has litigation" banner, but that banner is keyed to the worldwide family (priority 2013-03-12), not to this specific patent.

What the litigation activity on that family actually involves:

• Cerence Operating Co. v. Samsung Electronics Co., Ltd. et al., No. 2:23-cv-00482 (E.D. Tex., Marshall Div., Judge Gilstrap). Cerence's first Samsung complaint asserted, among others, US 11,393,461 and US 11,087,750 — both titled "Methods and apparatus for detecting a voice command" and both sharing the 2013-03-12 family lineage of the patent in question. A later report references a $78.5 million E.D. Tex. verdict for Cerence against Samsung, with post-trial briefing ongoing in 2026.
• A second Samsung complaint (filed 2024-03-15) asserted four Cerence patents; Samsung counterclaimed on US 10,395,657; 10,720,162; 11,823,682; and 9,583,103. Trial was scheduled for January 2026.
• In re Google LLC, No. 25-144 (Fed. Cir.). Samsung and Google jointly petitioned for mandamus after the PTAB (via the Acting Director's delegee) denied IPR institution on US 11,393,461 B2 and US 11,087,750 B2 under NHK-Fintiv, given parallel district-court progress. The Federal Circuit denied the mandamus petition (order filed 2025-11-06). The court did not reach validity on the merits. Related PTAB matters include IPR2024-01464/-01465 (Google), IPR2025-00457/-00458/-00459/-00460 (Samsung/Google), and IPR2024-01267 (Samsung, US 9,026,428 — terminated on settlement after institution, no merits decision).
• Cerence v. Amazon. On 2026-05-05 Cerence filed an ITC Section 337 complaint and parallel E.D. Tex. actions (No. 2:26-cv-00372, Judge Gilstrap; consolidated). The asserted patents are US 8,306,815; 9,203,972; 8,320,575; 11,929,073; and 8,355,484 — not 12,548,565.
• Cerence v. Sony Group (No. 2:25-cv-00757, E.D. Tex.) asserts US 7,840,579; 7,894,598; 8,189,810; and 9,171,541 — again not 12,548,565.

Family members I could correlate (same title, priority 2013-03-12, Cerence): US 11,393,461 B2 (app. 13/795,933), US 11,087,750 B2 (app. 15/238,238), US 11,676,600 B2, and US 12,548,565 B2 (app. 17/867,091). I could not trace the precise parent/continuation chain into 17/867,091 from the sources available to me.

---

6. Explicit uncertainties

1. Claim text not verified. I do not have the issued claims of US 12,548,565. §4 describes invention categories taken from the patent's own summary, which in this document conspicuously mirrors claim language ("Some embodiments include a method comprising…") — but I will not represent that this is the granted claim set. Independent-claim count, numbering, dependency structure, and any claim amendments made during prosecution are unknown to me.
2. Priority/continuity chain. The record states a 2013-03-12 earliest priority for a 2022-07-18 filing. I could not independently confirm the specific intermediate applications. One third-party aggregator lists a related publication (US 2021/0375285 A1) with a priority date of 2013-03-11 — a one-day discrepancy I am reporting as-is rather than correcting.
3. PTA and expiration. The 2034-02-22 adjusted-expiration date implies ~347 days of patent term adjustment. I have not verified the PTA certificate or any terminal disclaimer.
4. Litigation linkage is family-level, not patent-level. The litigation described in §5 involves sibling patents, not US 12,548,565 as far as I could determine.
5. Date inconsistency in sources. Several retrieved items (the Amazon ITC filing, the consolidated E.D. Tex. docket entries, and the Cerence jury-verdict coverage) are dated May–June 2026, which is later than the "current date" of April 26, 2026 given to me in this task. I am reporting them as retrieved, flagged here rather than silently reconciled.
6. No authoritative claim chart. No infringement contention, IPR petition, or claim-construction order naming US 12,548,565 was found, so I cannot tell you how Cerence or any challenger has characterized this specific patent's claims.

Bottom line: US 12,548,565 B2 is a Cerence continuation in the heavily litigated 2013-03-12 "detecting a voice command" family, claiming power-tiered always-on voice command detection using staged processing and contextual cues. It is not the number asserted in the Cerence v. Samsung, Cerence v. Google, Cerence v. Sony, or Cerence v. Amazon matters identified above — those target sibling patents.

I'll search for litigation involving US patent 12548565.

Let me search additional litigation-specific sources.

I found related Cerence litigation. Let me now determine whether US12548565 specifically appears in any of these cases.

Litigation Report: U.S. Patent No. 12,548,565

Bottom line

I did not find any litigation in which U.S. Patent No. 12,548,565 ("the '565 patent") is specifically identified as an asserted patent. I searched Google Patents, Unified Patents' portal, Justia/PACER mirrors, CourtListener, and Law360/news sources for the exact number. No complaint, ITC complaint, or PTAB filing naming the '565 patent surfaced.

I want to be explicit rather than overstate: the absence of results is not proof that no case exists, but I have no verified case to report. Anyone relying on this should independently confirm via PACER/Docket Navigator before treating it as complete.

Why no litigation is expected (yet)

• The '565 patent issued February 10, 2026 (application 17/867,091, filed July 18, 2022; priority claim March 12, 2013). It is roughly two months old as of the stated current date.
• It is a continuation in the Cerence Operating Company family (inventors Ganong, van Mulbregt, Sejnoha, Wilson) directed to "always-on" voice command detection including low-power-mode operation.
• A ~2-month-old patent rarely appears in public infringement dockets or PTAB filings, which is consistent with the null result.

The one relevant flag — and its important limitation

The Google Patents record for US12548565B2 displays a "Family has litigation" indicator linking to a Darts-IP family page (family ID 51529448), stating "First worldwide family litigation filed." Read literally, this signals that some member of the '565 patent's patent family has been litigated — not that the '565 patent itself has been asserted. Google Patents does not identify which family member or which case, and I could not resolve that link to a specific case within my search budget. Treat this as a lead, not a confirmed hit.

Closely related Cerence campaigns (NOT confirmed to assert the '565 patent)

Cerence and its affiliates are active litigants in voice-command/low-power voice detection technology, and some of these cases involve other patents sharing the '565 patent's March 12, 2013 priority family. I found these, but I could not verify that any of them names the '565 patent:

Additional context: PTAB proceedings exist against other, older Cerence patents (e.g., IPR2025-00458/00459/00460 on U.S. 8,682,671; 8,825,486; 8,914,291), in which institution was denied under § 314(a). None of those patents is the '565 patent.

Method and caveats

• Searched exact strings "12548565," "US12548565," and "US 12,548,565" against patent-litigation aggregators. Many results returned were coincidental matches to a different number or parameter (e.g., Unified Patents' page for U.S. 7,846,754, whose application number happens to be 12,548,565; an Estonian business-registry code 12548565; a UK companies-registry entry). Per your instruction, I did not conflate these with the patent — the application number 12,548,565 on U.S. 7,846,754 is the Samsung LED patent, not the '565 patent, and I have flagged it as a distinct, coincidental match.
• Date note: several sources above are dated June–July 2026, which postdates the "current date" given to me (April 26, 2026). I am reporting them as retrieved rather than reconciling the discrepancy, and you should verify their currency.
• I could not complete verification of the specific patent numbers asserted in the Cerence v. Amazon district-court complaint and ITC complaint before exhausting my search steps. Those complaints' exhibit lists would be the authoritative place to determine whether the '565 patent is asserted.

Recommendation: Pull the complaint and Exhibit 1–13 lists in E.D. Tex. 2:26-cv-00372 and the ITC Section 337 complaint against Amazon, and run an exact-number query in Docket Navigator or PACER. Those are the most likely places a newly issued Cerence continuation like the '565 patent would first appear.

I'll start by verifying the structured data against what's publicly findable, checking both this patent and its family (the priority-2013 Nuance/Cerence "detecting a voice command" line).

Family map is emerging: the '565 patent is a direct continuation in the Nuance/Cerence "detecting a voice command" family, and IPRs were filed against sibling patents US 11,087,750 and US 11,393,461. Let me pin down those proceedings and check for anything against '565 or its immediate parent '600.

Proceedings overview

There are zero AIA trial proceedings on file against US 12,548,565 — the USPTO ODP structured data returns none, and my web search surfaced none; the only PTAB activity in this patent family is against two sibling patents in the same 2013-03-12 priority chain (US 11,393,461 and US 11,087,750), and both of those petitions were denied institution, so no claim of any family member has been canceled and the defensive posture is: the family has never been tested on the merits at the PTAB — there is no invalidation ammunition to borrow, but also no estoppel cloud and a fully open prior-art field.

Critical framing. The proceedings below are not proceedings on US 12,548,565. They are against US 11,393,461 B2 ('461) and US 11,087,750 B2 ('750), which share the '565 patent's specification, inventors (Ganong III, van Mulbregt, Sejnoha, Wilson) and 2013-03-12 priority date, and which were asserted against Samsung in the same E.D. Tex. case. I present them because they are the closest thing to a PTAB track record for this asset — but you cannot tell a judge that "claims of '565 were canceled," because none were.

---

IPR2024-01465 — Google LLC, Samsung Electronics Co., Ltd. & [Samsung Electronics America, Inc.](/litigations/by-plaintiff/Samsung%20Electronics%20America%2C%20Inc.) v. Cerence Operating Company

• Patent at issue: US 11,087,750 B2 ("Methods and apparatus for detecting a voice command") — not US 12,548,565
• Type: Inter Partes Review
• Filed: 2024-10-17
• Status: Institution denied — Board denied institution on 2025-04-23; request for Director Review denied; Federal Circuit mandamus denied 2025-11-06. Plain English: the PTAB never reached the merits; the patent has not been upheld on the merits, it simply was never reviewed.
• Judge panel: The institution-denial panel composition is not confirmed in the sources I could reach. The consolidated Director Review order covering both -01464 and -01465 was issued by Senior Lead Administrative Patent Judge Michelle N. Ankenbrand, performing the duties of Director Review Executive (Acting Under Secretary/Acting Director Coke Morgan Stewart was recused — see Notice of Delegation, Paper 18). Korean press coverage attributes the underlying denial to APJ "Michelle N. Wormmeester"; I flag the name discrepancy rather than resolve it. (PTAB E2E record, IPR2024-01464/-01465)
• Petition grounds: Challenged claims 1–19, all under 35 U.S.C. § 103:
  • Ground 1 — claims 1–8, 11, 13–15, 19 obvious over Basye (US 2014/0163978) in view of Shenhav (US 2014/0006825)
  • Ground 2 — claims 9, 16 over Basye/Shenhav + Parker (US 2015/0106089)
  • Ground 3 — claims 10, 18 over Basye/Shenhav + LeBeau (US 2012/0034904)
  • Additional ground — claims 12, 17 over Basye/Shenhav + Conroy (US 2011/0314185)
  • Petitioner also argued against § 325(d) discretionary denial, asserting Basye/Shenhav/Parker/LeBeau/Conroy were not substantively before the Examiner. (Grounds summary)
• Institution decision: Denied on 2025-04-23 under 35 U.S.C. § 314(a) (Fintiv). The Board credited Patent Owner's showing that the parallel E.D. Tex. trial date fell "almost seven months before the final written decision in this proceeding would be due," and that "concerns of inefficiency and the possibility of conflicting decisions" weighed against institution because Samsung's district-court invalidity contentions contemplated combining Apple's Siri and Samsung's S Voice with the same prior art cited in the petitions. The Board applied Chief APJ Boalick's 2025-03-24 Guidance Memo rescinding the Vidal 2022 Fintiv memo (retroactively, including to petitions already filed, such as this one) — meaning Samsung's Sotera stipulation was reduced from dispositive to "part of the holistic analysis." (Petitioners separately sought leave for supplemental Fintiv briefing on 2025-04-04.)
• Final Written Decision: None — no FWD exists. No claim of '750 was canceled or confirmed.
• Settlement / termination: No settlement. Terminated by institution denial, not by party agreement.
• Appeal: No Federal Circuit appeal is possible — § 314(d) makes institution decisions "final and nonappealable." Petitioners instead sought mandamus: In re Google LLC, Samsung Electronics Co., Ltd., Samsung Electronics America, Inc., No. 2025-144 (Fed. Cir.), petition filed 2025-08-18, denied 2025-11-06 (nonprecedential order; Dyk, Linn, Stoll, JJ.; authored by Linn). The court held mandamus is "ordinarily unavailable for review of institution decisions — including decisions denying institution based on the progress of parallel district court proceedings involving the same patents," citing Mylan v. Janssen, and noted only possible exceptions for "colorable constitutional claims." (Fed. Cir. order; mandamus petition)
• Defensive value: You cannot cite this as invalidating anything. What it does give you is a proven discretionary-denial template: the Board has already refused to institute on this family where an E.D. Tex. case has an earlier trial date, even against a petitioner (Google) who was not a party to the district court case and even with a Sotera stipulation on file. If you intend to file an IPR against '565 while a parallel case is pending, budget for a Fintiv denial and plan a compelling-merits record from day one.

IPR2024-01464 — Google LLC, Samsung Electronics Co., Ltd. & Samsung Electronics America, Inc. v. Cerence Operating Company

• Patent at issue: US 11,393,461 B2 ("Methods and apparatus for detecting a voice command") — not US 12,548,565
• Type: Inter Partes Review
• Filed: Filed jointly with IPR2024-01465; the precise -01464 filing date was not independently confirmed (companion petition dated 2024-10-17)
• Status: Institution denied (2025-04-23); Director Review denied (consolidated order covering -01464 and -01465); mandamus denied 2025-11-06
• Judge panel: Same posture as -01465 — Director Review order by Senior Lead APJ Ankenbrand; underlying denial panel not confirmed
• Petition grounds: Not verified in the sources I retrieved; Petitioners' consolidated papers confirm the grounds raised were § 103 obviousness grounds and that Samsung's district-court contentions involved Siri and S Voice combinations with art overlapping the petitions. Do not attribute specific references to -01464 without pulling the petition. (Petition/record)
• Institution decision: Denied 2025-04-23 on § 314(a)/Fintiv grounds, on the same reasoning as -01465 (Board "denied institution of the petitions challenging the '461 and '750 patents purely on Fintiv grounds")
• Final Written Decision: None — no FWD exists.
• Settlement / termination: None; terminated by institution denial
• Appeal: Same No. 2025-144 mandamus route; denied 2025-11-06. Google Patents' legal-events record for the '461 patent confirms the filing: "IPR | AIA trial proceeding filed... TRIAL NO: IPR2024-01464 | Opponent name: GOOGLE LLC, SAMSUNG ELECTRONICS CO., LTD" (source)
• Defensive value: Same as -01465 — no claim-level outcome to leverage. The '461 is the sibling closest in disclosure to '565, so this is the best proxy evidence of how the Board has treated the family's substance procedurally (i.e., it hasn't).

IPR2024-00978, IPR2024-00979, IPR2024-01267 — Samsung Electronics Co., Ltd. (and Google LLC) v. Cerence Operating Company

• Patent(s) at issue: Not verified. These three derive from the same five-patent E.D. Tex. assertion campaign but are not the '461 or '750 (those were denied). I could not confirm which patents they name — treat the patent numbers as unverified.
• Type: Inter Partes Review
• Filed: Filed approximately two months before -01464/-01465 (i.e., on or around 2024-08)
• Status: Instituted — IPR2024-00978 instituted by Paper dated 2025-02-13; IPR2024-00979 and IPR2024-01267 instituted by papers dated 2025-01-31. Samsung's Sotera stipulations were dispositive in each, under the pre-rescission Vidal framework.
• Judge panel: Not confirmed
• Final Written Decision: Not confirmed in the sources retrieved — I cannot tell you whether these have reached FWD, settled, or are still in trial. Do not assume.
• Defensive value: These instituted proceedings are the reason -01464/-01465 are so instructive — the very same petitioners, on the very same patents-in-suit, got in on three petitions in January–February 2025 and were then kept out on the two filed against the voice-command patents after the March 2025 rescission. The turning point is the change in discretionary-denial policy, not the merits. (Citations at p. 3, n.4 of the mandamus petition)

Related appellate proceeding — In re Google LLC, Samsung Electronics Co., Ltd., Samsung Electronics America, Inc., No. 2025-144 (Fed. Cir.)

• Type: Petition for writ of mandamus to the USPTO, Nos. IPR2024-01464 and -01465
• Filed: 2025-08-18 · Disposition: Denied 2025-11-06 (nonprecedential; Dyk, Linn, Stoll, JJ.)
• Issues: (1) whether the retroactive application of the 2025 rescission of the 2022 Vidal Fintiv memo violated due process and the APA; (2) whether the Board's demand that petitioners surrender defenses broader than statutory IPR estoppel was ultra vires; (3) whether mandamus is available notwithstanding Mylan
• CourtListener / source: Fed. Cir. order (PDF)
• Defensive value: This closes the door on appellate relief from a Fintiv-based denial. If your IPR against '565 is denied on parallel-litigation grounds, there is no meaningful appellate remedy — plan accordingly at the petition stage.

---

Strategic summary

Claim status. No claim of US 12,548,565 has been CANCELED, and — strictly speaking — no claim has been SUSTAINED either, because no AIA trial on '565 (or on its immediate parent, US 11,676,600) has ever been instituted. Every claim of '565 is UNTESTED at the PTAB. The same is true of the sibling '461 and '750: the Board denied institution on discretionary grounds without reaching § 102 or § 103. I was not able to verify the claim set of '565 from the sources retrieved, so I am deliberately not characterizing individual claim numbers — anyone telling you "the independent claim is the contextual-cue claim" is inferring from the shared abstract, not reading the patent. Pull the granted claims from Patent Center before you build anything on them.

Estoppel landscape. This is the most favorable fact in the file for a defendant: there is no IPR estoppel anywhere in this family. Section 315(e)(2) estoppel attaches only where an IPR "results in a final written decision under section 318(a)" — and there is no FWD in -01464 or -01465. Samsung's Sotera-style stipulation was expressly conditioned on institution ("if the PTAB institutes review"), so it never ripened and imposes no constraint. Every prior-art ground remains available in district court, including the Siri/S Voice + prior-art combinations Samsung advanced in E.D. Tex. Practically: you may raise any § 102 or § 103 ground you can find, in any forum, with no estoppel shadow. The only caveat is the ordinary § 315(b) one-year clock running from service of a complaint asserting '565 — and note that whether prior service of a complaint asserting a family member starts that clock for '565 is a question that needs its own analysis; I am not asserting it does.

Pattern signals. (1) The same petitioner group — Google + Samsung, jointly — filed on this family, with Samsung as the district-court defendant and Google as a non-party co-petitioner (a deliberate strategy to escape Fintiv; it failed anyway). (2) The patent owner, Cerence Operating Company, litigated the discretionary-denial issue hard and won at every level — Patent Owner Preliminary Responses, supplemental Fintiv briefing, Director Review, and the Federal Circuit. Cerence has not had to defend a single claim on the merits at the PTAB. (3) No defensive aggregator (Unified Patents or similar) appears anywhere in the chain; the only Unified presence is an analytics portal page for US 11,676,600, which is a database entry, not a filing. (4) The controlling variable across the whole family is policy, not art: the January–February 2025 institutions (IPR2024-00978/-00979/-01267) versus the April 2025 denials (-01464/-01465) turn on the rescission of the Vidal memo and the March 2025 workflow-management regime, under which, as petitioners themselves put it, "all but two of seventy Director decisions were discretionary denials."

---

Recommended next steps

1. Say it plainly for the record: there is no PTAB proceeding on US 12,548,565. No FWD, no canceled claims, nothing to cite, nothing to distinguish. If a demand letter asserts '565, you cannot answer "those claims are already dead" — no tribunal has ever assessed them.
2. Watch the PGR window, which closes 2026-11-10. '565 granted 2026-02-10; a post-grant review petition must be filed within nine months of grant, i.e., by 2026-11-10. Today is 2026-10-01 — roughly 40 days remain. PGR is the only vehicle that reaches § 101 and § 112 grounds, plus any § 102/§ 103 art (including art that a petitioner might want to pair with public-use or system-prior-art evidence that IPR cannot reach). If you have eligibility or written-description/Enablement theories on the "contextual cue" subject matter, this is the deadline that matters. Confirm the statutory date against the face of the patent and the ODP record before relying on it.
3. If you intend an IPR, do not file a garden-variety Fintiv-exposed petition. The Board has already declined to institute on this family in a parallel-litigation posture, the Director Review request was denied, and the Federal Circuit refused mandamus on 2025-11-06. Build a compelling-merits record, consider a stipulation broader than Sotera (the Board's post-rescission position, per petitioners' own account, is that Sotera alone is insufficient and that a petitioner may need to surrender grounds beyond statutory estoppel), and if there is no parallel case — as with Google — argue that factor affirmatively. Note the Board's practice as of March 2026 (Director Squires) continues to include discretionary denials on these grounds.
4. Mine the sibling petitions for a ready-made roadmap — with your eyes open. Basye (US 2014/0163978), Shenhav (US 2014/0006825), Parker (US 2015/0106089), LeBeau (US 2012/0034904) and Conroy (US 2011/0314185) are all pre-2013-03-12 art mapped against the shared specification. Cerence has already seen and briefed every one of them, so treat them as a floor, not a ceiling — and note that '565's later continuation filing date (2022-07-18) does not move the effective priority date, which is the family's 2013-03-12 date.
5. Contact the PTAB-relevant record directly to close my gaps. The items I could not verify and would not assert: (a) the claim set of '565 and its effective filing/priority chain on the face of the patent; (b) the identity of the patents in IPR2024-00978/-00979/-01267 and whether any have reached FWD; (c) the panel composition of the -01464/-01465 institution denials. Pull these from PTAB E2E, the Google Patents legal-events tab for the '461, and Patent Center for '565 before you file anything that depends on them.

I'll research the assignment record for US 12,548,565, starting with the legal events already captured in the patent text and then verifying against assignment and corporate sources.

Let me dig into the specific assignment records and the Cerence/Nuance corporate history.

Let me pin down the specific reel/frame records and the Cerence corporate/credit-facility structure.

Assignment & Ownership Chain Report — U.S. Patent No. 12,548,565

Analyst's note on sourcing. I was able to retrieve the patent's legal-events record (mirrored from USPTO assignment data) and corroborating assignment-index pages, but my search budget was exhausted before I could open the Assignment Center rows themselves. I therefore report reel/frame numbers only where they are actually in the record I retrieved, and I mark everything else "not retrieved" rather than reconstructing it. The one reel/frame pair I have — 067417/0303 — comes from the release entry's own description text. Treat this as an intermediate work product requiring confirmation at USPTO Assignment Center (search by patent number 12548565 or application 17/867,091) before reliance.

---

Inventors

Basis for the employer attribution. Cerence did not exist as an entity until the Nuance spin-off closed on 2019-10-01, so no inventor could have been a Cerence employee at the 2013-03-12 priority date. The application was filed 2022-07-18 (App. 17/867,091) with Cerence Operating Company as applicant of record, and the family's U.S. rights reached Cerence through the Nuance→Cerence spin-off IP assignment (see below). That chain is what makes the "Cerence employee at continuation filing" column safe; the Nuance attribution is an inference from entity-existence timing plus the family's provenance, not from a located employment record.

Pattern observations.

• No inventor-departure data located. I found no evidence of any inventor leaving the original assignee within 12 months of filing — but I also found no current-assignment or employment records either way. I am not going to infer a fire-sale precursor from silence. Unclear/none.
• The more notable pattern is continuation-farming, not inventor attrition. The same four-inventor set persists across a 2013 priority filing and a 2022 continuation, with issuance in 2026 — a ~13-year pendency chain. That is a term/claim-scope management strategy typical of a portfolio that is being curated for assertion and licensing. It is not, by itself, an NPE signal.
• Subject-matter mismatch worth noting: the inventors' output here is mobile-device, low-power "always-on" voice trigger technology, yet it landed in an automotive spin-off. That is a function of how the 2019 Nuance separation split the business units, not a transfer to an unrelated party.

---

Original assignee

Cerence Operating Company (the entity named as original and current assignee on the Google Patents face of the '565 patent).

• Relationship to the public parent: Cerence Operating Company is the U.S. operating/patent-holding subsidiary of Cerence Inc. (NASDAQ: CRNC), the entity created by Nuance Communications, Inc.'s tax-free spin-off of its automotive business, effective 2019-10-01 (distribution ratio 1 CRNC share per 8 NUAN shares; record date 2019-09-17; regular-way trading from 2019-10-02).
• Primary line of business: automotive/mobility AI — in-car voice assistants, natural-language understanding, voice biometrics, gesture/gaze. Cerence historically powered voice AI in roughly 280–300 million vehicles across 70+ languages, supplying nearly every major automaker (Audi, BMW, Daimler, Ford, Geely, GM, SAIC, Toyota, et al.).
• Did it ship a product embodying the claims? Yes, in the ordinary sense — Cerence's in-car and cloud voice stack is the commercial embodiment, and the same entity is the named plaintiff in Cerence's infringement suits. Note that the claim subject matter (mobile device in a low-power mode detecting a voice command without an explicit trigger) is a handset-facing articulation of that in-car stack.
• Current status: operating and publicly traded. I found no Chapter 7/11 filing for Cerence Inc. or Cerence Operating Company. Cerence has separately been under financial pressure in the mid-2020s (revenue concentration in a few OEMs), and it publicly explored strategic alternatives; I did not verify those developments against primary filings, so I flag them as unverified context only.

---

Assignment timeline

Three links are traceable: two visible directly on the '565 patent's legal-events record, plus the upstream spin-off transfer. I am explicitly not listing a 2013 inventors→Nuance assignment, because I did not retrieve that record — see the gap note after the table.

1. Executed on/about 2019-09-30 / recorded not retrieved — Reel/Frame not retrieved

• Conveyance: Assignment (bulk intellectual-property assignment executed under the Spin-Off's Intellectual Property Agreement, which expressly required recordation of Patent Assignment Agreements per its § 2.01)
• Assignor: Nuance Communications, Inc.
• Assignee: Cerence Operating Company (with Cerence Inc. as the spin-off parent)
• Correspondent: not retrieved. (If this recurs across Cerence's recorded assignments, it is very likely Nuance/Cerence in-house IP counsel — the same signing pattern appears on Nuance-family recordings, e.g., David Greenbaum, VP, Litigation & IP, on the 2021-01-01 Nuance→Voice Inventions assignment, and Leanne J. Fitzgerald, Associate General Counsel – IP, Nuance Communications, Inc., on an earlier Dictaphone/Nuance recording. I did not confirm either as correspondent of record here.)
• Context: internal corporate reorganization (spin-off carve-out), not a monetization transfer.
• Corroboration: indexed as "Patent Assignment from Nuance Communications Inc. to Cerence Operating Company" — https://www.plainsite.org/patents/assignment.html?id=[9281678](/patent/9281678)
• Scope caveat: the '565 application itself (17/867,091) was filed in 2022, after this transfer. This link therefore covers the '565 patent's ancestor application in the priority chain, not the '565 application directly.

2. Executed date not retrieved / recorded 2024-04-15 — Reel/Frame 067417/0303

• Conveyance: Security Agreement (notice of grant of security interest in patents)
• Assignor: Cerence Operating Company
• Assignee: Wells Fargo Bank, N.A., as Collateral Agent
• Correspondent: not retrieved — this is the single most important field I could not capture (see NPE signal 3).
• Context: securitization. This is pledge/perfection of the patent as collateral under a secured credit facility, not a sale. Cerence's collateral agreements of this vintage grant the Collateral Agent rights, title and interest in "all Patents of the United States of America, including those listed on Schedule I," with sale-of-collateral remedies on default. A parallel recording covering a broad Cerence patent portfolio (7,987,092; 8,370,152; 9,418,673; 7,379,872; 7,702,503; 7,412,377; 7,275,030; 7,487,451; 7,739,600; 7,634,412, and others) is indexed at https://www.plainsite.org/patents/assignment.html?id=[10412757](/patent/10412757) ("Patent Assignment from Cerence Operating Company to Wells Fargo Bank, NA").
• Reading caveat: Google Patents shows no reel/frame on the 2024-04-15 row itself; 067417/0303 is taken from the 2025-01-02 release row's citation of the security interest it releases. If the release instead reaches back to an earlier Wells Fargo recording, then the April 2024 security agreement sits at a different, unretrieved reel/frame. Confirm before quoting.

3. Executed date not retrieved / recorded 2025-01-02 — Reel/Frame of this release not retrieved (it releases Reel 067417/0303)

• Conveyance: Release (release of security interest in patents previously recorded at Reel 067417 / Frame 0303)
• Assignor: Wells Fargo Bank, National Association
• Assignee: Cerence Operating Company
• Correspondent: not retrieved.
• Context: securitization unwind — release of the collateral pledge, consistent with payoff, refinancing, or an amend-and-restate that superseded the prior grant notices. No ownership change. Cerence Operating Company remains assignee of record before and after.

Gap note — records I expected but did not retrieve

• c. 2013 inventors→Nuance Communications assignment. Expected, because the priority filing predates Cerence's existence by six years. Not retrieved; no reel/frame asserted.
• 2019–2020 Nuance→Cerence case-specific assignments. The IP Agreement contemplated "such additional case specific assignments as deemed appropriate or necessary." Individual applications (including the '565 ancestor) were likely recorded separately or by schedule. Not retrieved.
• A possible intermediate Cerence Inc.→Cerence Operating Company transfer if title sat at the parent before being pushed down to the operating subsidiary. Not retrieved; I have no evidence either way.

If the Assignment Center in fact shows no records for 17/867,091/12,548,565, then the two legal events above are being surfaced by Google Patents from the portfolio-level recordings rather than from case-level rows — which would mean Cerence holds title through a schedule-based recording. Either way, the ownership conclusion is unchanged: Cerence Operating Company is the owner of record.

---

Timeline diagram

timeline
    title Ownership of US 12548565
    2013 : Priority application filed by Nuance
    2019 : Nuance spins off Cerence
         : IP assigned to Cerence Operating Co
    2022 : Continuation 17 867 091 filed by Cerence
    2024 : Security agreement to Wells Fargo
    2025 : Security interest released to Cerence
    2026 : Patent granted to Cerence Operating Co

---

NPE / troll-pattern signals

1. Shell-entity transfer — NOT PRESENT.
Every assignee in the chain is either Cerence Operating Company (the operating and litigating subsidiary of a NASDAQ-listed company) or Wells Fargo Bank, N.A. as Collateral Agent (Reel 067417/0303, recorded 2024-04-15). No "IP / Patents / Licensing / Holdings / Ventures" suffix appears. No single-purpose Delaware or Texas LLC appears. The Wells Fargo entry is a secured-lending artifact with a stated collateral-agent capacity, which is the opposite of an anonymous shell: it names a regulated bank in a fiduciary capacity and is released 9 months later (2025-01-02).

2. Known asserter in the chain — NOT PRESENT (adjacent context only).
Neither Cerence Operating Company nor Wells Fargo matches any entity on the Acacia, Marathon, Intellectual Ventures, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation, or Spangenberg lists, nor any Unified Patents/RPX high-frequency-plaintiff directory I checked. Cerence does assert patents, but it asserts as the operating company that sells the accused-adjacent technology — that is competitor litigation, not NPE assertion.
Adjacent context worth recording: the former parent, Nuance Communications, Inc., assigned a separate natural-language-speech patent portfolio to an entity identified in the recording as "Voice Inventions," effective 2021-01-01 (signed by David Greenbaum, VP, Litigation & IP; portfolio includes 7,398,209; 7,502,738; 7,809,570; 8,015,006; 8,112,275; 8,140,327; 8,155,962; 8,731,929; 9,734,825; 7,693,720; 9,031,845). That evidences a monetization posture inside the former parent, and it matters because those patents are cited as prior art against Cerence-family claims — but it is not in the '565 chain and must not be counted as a chain signal. I did not verify litigation activity by that assignee.

3. Repeat correspondent across the chain — NOT DETERMINABLE.
I could not retrieve the correspondent of record for any link. This is the signal designed to survive shell naming, and its absence is a data gap, not a clean result. Two names are worth checking against the Assignment Center rows: David Greenbaum (VP, Litigation & IP) and Leanne J. Fitzgerald (Associate General Counsel – IP), both of whom execute Nuance/Cerence-family IP assignments and would be the natural correspondents if the recordings were made in-house. A single appearance is not a finding; the test is recurrence across Reel 067417/0303 and the 2025-01-02 release. Re-run this query.

4. Cascading transfers — NOT PRESENT.
The chain contains no LLC-to-LLC sequence. The only two post-2022 events are a security agreement (2024-04-15) and its release (2025-01-02) — 9 months apart, assignee identical throughout (Cerence Operating Company). No shared registered-agent addresses, no common principals, no rapid churn.

5. Pre-litigation transfer — NOT PRESENT.
Cerence Operating Company is both the owner of record and the plaintiff in its suits, so there is no "clean up title, then sue" transfer to evaluate. Timing check against the litigation record: security agreement 2024-04-15 → Apple suit 2025-09-04 (~17 months); release 2025-01-02 → Apple suit (~8 months); → Amazon suits 2026-05-04 (>16 months). The release is the closest event to a filing, and it is 8 months out and is not an ownership transfer. Outside the 6-month window.

6. Bankruptcy fire-sale — NOT PRESENT.
No Chapter 7/11 filing by Cerence Inc. or Cerence Operating Company located. Cerence Inc. remained a going concern with public reporting through the relevant period. The 2025-01-02 release is more consistent with an ordinary facility payoff/amendment than with a distressed asset sale, and in any event a release returns the collateral to the owner rather than divesting it.

7. Privateering — NOT PRESENT.
No transfer from Cerence to an NPE asserting on Cerence's behalf. Cerence sues in its own name (Cerence Operating Co. v. Apple, W.D. Tex. 7:25-cv-00400; Cerence Operating Co. v. Samsung, E.D. Tex. 2:24-cv-00181; Cerence Operating Co. v. Amazon, E.D. Tex. 2:26-cv-00372/00373), and the family's Darts-IP "Family has litigation" flag is explained by that direct-assertion pattern.

8. Defensive aggregator — NOT PRESENT.
The chain terminates at Cerence Operating Company. No RPX, AST, LOT, Unified Patents, or OIN acquisition. The patent is live assertion ammunition, not neutralized.

---

Verdict

Operating-company assertion.

Cerence Operating Company — the operating subsidiary of NASDAQ-listed Cerence Inc. — is the owner of record both at filing (2022-07-18) and today, and the only recorded encumbrances are a security agreement to Wells Fargo Bank, N.A. as Collateral Agent (recorded 2024-04-15, Reel 067417/0303) and its release (recorded 2025-01-02); neither is an ownership transfer, and no shell LLC, known NPE, or defensive aggregator appears anywhere in the chain. The same entity ships the accused-adjacent voice product and sues actual competitors (Apple, Samsung, Amazon), which is the definition of operating-company assertion rather than NPE conduct, even though Cerence is a notably active and portfolio-curating litigant whose family carries a Darts-IP litigation flag.

Confidence: moderate-to-high on the character of the chain (security interest + release, both to/from a named bank and the operating subsidiary); low on completeness, because I could not retrieve the correspondent of record on any link and could not independently confirm the reel/frame for the 2019 spin-off assignment or for the 2025 release itself. The two fields most likely to change this verdict if filled in are (i) a repeated NPE-side correspondent across Reel 067417/0303 and the release, and (ii) confirmation that no case-level assignment ever moved title off Cerence Operating Company.

Verification: query patent number 12548565 or application 17/867,091 at https://assignmentcenter.uspto.gov/ (mirror: https://assignment.uspto.gov/patent/index.html), then pull the reel/frame behind the 2024-04-15 and 2025-01-02 rows, and cross-check the October 2019 Nuance→Cerence recordings.

---

Contradictions / discrepancies flagged

1. Date basis conflict with the previously generated litigation section. That section assumed a "current date" of 2026-04-26 and called the patent "roughly two months old." The operating date supplied for this task is 2026-10-01, making the '565 patent (granted 2026-02-10) roughly eight months old. This shifts every timing calculation in signal 5. Under either date, the pre-litigation-transfer conclusion holds, but the litigation section's own caveat — that it was reporting June–July 2026 sources which postdate its assumed present — is the same discrepancy, and one of the two date anchors is wrong.
2. Reel/frame attribution. Reel 067417/0303 is cited in the record as the security interest being released on 2025-01-02. I have attributed it to the 2024-04-15 security agreement by elimination, since no other security agreement appears in the '565 record. If the Assignment Center shows the April 2024 grant at a different reel/frame, the attribution above needs correcting — this is the single most fragile point in this report.
3. Term arithmetic (not an ownership issue, but worth correcting before anyone misread it). The "adjusted expiration 2034-02-22" is consistent with a 20-year term from the 2013-03-12 priority date plus roughly 347 days of patent term adjustment; it is not evidence of a 2014 filing date.
4. No conflict found between this analysis and the litigation section's ownership statements (original and current assignee = Cerence Operating Co.), or with its ranking of the Cerence v. Amazon / ITC matters as the most likely venues for a newly issued Cerence continuation to surface.

I'll search for the specific patent's citation records and prior art references.

The first searches returned mostly unrelated material. Let me target the patent's own front-page citation record specifically.

Let me try to retrieve the actual front-page citation list from the patent record itself.

Prior-Art Analysis — US 12,548,565 B2 ("Methods and apparatus for detecting a voice command")

⚠️ Critical threshold finding: the patent's own citation record is not in the authoritative document

Before anything else, I must flag a limitation that materially constrains this task, consistent with the rule against fabrication:

The authoritative full text you supplied for US 12,548,565 B2 is truncated before the claims and does not include the front-page "(56) References Cited" section. Google Patents pages normally render a References Cited block (U.S. patent documents, foreign patent documents, and "Other Publications"), plus a Cited By block. Neither is present in the text I was given, and my retrieval budget did not surface the verbatim (56) list for this specific number.

Consequence: I cannot truthfully present "each patent citation for 12,548,565" as an enumerated, verified list, and I will not construct one from memory. What I can do is (a) report what the record does establish about the '565 patent's pre-grant citation lineage, (b) supply the closest verifiable citation data from the patent's own continuation family, and (c) identify the prior art that has actually been arrayed against this family in adversarial proceedings. Each item below is labeled with its verification status.

---

1. What the record does establish

1a. Family linkage (not prior art — flagged to prevent conflation)

Multiple Google Patents pages list the following as co-family documents sharing the title "Methods and apparatus for detecting a voice command" and priority 2013-03-12:

Source: Google Patents family listings (e.g., patents.google.com/patent/US11676600B2/en; patents.google.com/patent/US20150106085A1/en).

*Note the recurring one-day priority discrepancy: Unified Patents' page for US 11,676,600 B2 records the priority date as 2013-03-11, while Google Patents records 2013-03-12 for the family. Reported as-is, not reconciled.

These are family members, NOT prior art under §102. They cannot anticipate the '565 patent (same inventive entity/family, common priority). Any analysis that treats them as §102 art would be wrong. I include them only to correct a common error and to explain why the citation lists of the siblings are the best available proxy for the '565 list.

1b. Verified citation-volume data (closest available proxy)

Unified Patents' record for sibling US 11,676,600 B2 (same specification, same inventors, same priority) states:

• Patent Art: 95 references
• Non-Patent Literature: 8 references
• Referenced By: 0
• Examiner: Ho, Huy C.

Source: https://portal.unifiedpatents.com/patents/patent/US-11676600-B2

Because 17/397,522 ('600) and 17/867,091 ('565) are same-family continuations with the same examiner-facing specification, the '565 patent's examiner-cited art is highly likely to overlap substantially with this ~95-reference body. I have not verified the overlap, and I am explicitly labeling this as an inference, not a finding.

1c. A probable cited reference (low confidence)

A Google Patents citation-listing page surfaced US 2013/0080167 A1 — "Background Speech Recognition Assistant Using Speaker Verification" (priority 2013-03-12) in a table also containing US 12,548,565 B2 and US 11,676,600 B2. Source: https://patents.google.com/patent/US20130080167#13. The placement suggests a citation/cited-by relationship between the '565 family and the Mengibar-style background-speech-assistant disclosure, but I could not confirm whether '565 cites it or is cited by it. Treat as a lead only.

---

2. Prior art actually arrayed against this family in adversarial proceedings (verified)

This is the most concrete, verifiable prior-art evidence I can give you, and it is more probative than a front-page citation list because it reflects a third party's considered §103 grounds.

Proceeding: Inter partes review petitions challenging US 11,087,750 B2 (sibling; "the '750 Patent"), petition document at https://ptacts.uspto.gov/ptacts/public-informations/petitions/1556713/...

The petition asserts, verbatim from the record:

"Ground 2: Basye in Combination With Shenhav and Parker Renders Obvious Claims 9 and 16"
"A POSITA would have been motivated to combine the teachings of Basye, Shenhav, and Parker ('Basye-Shenhav-Parker')…"

Statutory posture: this is a §103 obviousness combination, not a §102 anticipation. Under the strict rule in your task ("which claim(s) it potentially anticipates under 35 U.S.C. § 102"), the honest answer is: none of these three were asserted as §102 anticipatory art in that record — none is alleged to disclose every limitation of a single claim. If you want §102 candidates, Basye is the only one with the structural breadth to be a plausible single-reference §102 reference, and even that was pleaded as §103.

Important caveat: the ground above targets the '750 Patent, not the '565 patent. As noted in the previously generated litigation section, IPR2024-01464/-01465 (Google) and IPR2025-00457/-00458/-00459/-00460 (Samsung/Google) on US 11,393,461 B2 and US 11,087,750 B2 were denied institution under §314(a)/NHK-Fintiv, and the Federal Circuit denied mandamus (In re Google LLC, No. 25-144). No merits decision on validity exists for any sibling, and I found no IPR naming 12,548,565.

---

3. Field references appearing in the same prosecution records (context, not mapped)

The sibling-patent PDFs and search-report filings I retrieved reference the following document families repeatedly in this art area. I am listing them as context only; I have not verified that any of them appears on the front page of US 12,548,565:

• WO 2013/163113 and WO 2014/008194, WO 2014/130463, WO 2015/025330 (foreign patent documents appearing in Cerence/Nuance file histories — e.g., US11276406.pdf front page).
• WO 2017/138934 A1 (Nuance, "Techniques for spatially selective wake-up word recognition and related systems and methods") — appears in the '461/'600 family citation neighborhood, along with its U.S. siblings US 11,437,020 B2, US 11,600,269 B2, US 11,545,146 B2.
• Non-patent literature common to this family's prosecution: Germain et al., "Speaker and Noise Independent Voice Activity Detection" (2013); Ramirez et al., "Statistical Voice Activity Detection Using a Multiple Observation Likelihood Ratio Test," IEEE SPL 12(10), Oct. 2005; Kwon et al., "Unsupervised Speaker Indexing Using Generic Models," IEEE Trans. Speech & Audio 13(5), 2005; "Speaker Recognition," Wikipedia.

---

4. §102 mapping — stated honestly

A §102 anticipation analysis requires (i) the verbatim claim text of the '565 patent and (ii) the reference's disclosure. I have neither verified claim text nor the reference set for '565. I therefore cannot provide a defensible "which claims does reference X anticipate" table, and any such table I produced would be fabrication. Two points I can state with confidence:

1. For any reference asserted only as part of a multi-reference §103 combination (as Basye/Shenhav/Parker were), the correct answer to "which claims does it anticipate under §102?" is none, as pleaded — combinations are §103 by definition.
2. The '565 patent's claims are continuation claims from a 2013-03-12 priority chain. Any reference you evaluate must be prior art as of the applicable 2013 critical date — a point the previously generated sections flagged regarding the 2013-03-11 vs. 2013-03-12 discrepancy. That one-day gap is potentially outcome-determinative for any reference published or filed on 2013-03-11/12, so it must be resolved (via the priority document itself) before any §102 determination.

---

5. Step-2 recommendations to close the gap

To give you the verified citation list the task asks for, pull these primary records:

1. USPTO Patent Center / Patent Public Search for application 17/867,091 → download the issued patent PDF's front page and the PTO-892 / IDS forms. This yields the verbatim (56) list.
2. Patent Public Search (ppubs.uspto.gov) → retrieve US 12,548,565 B2 full document, including References Cited.
3. Google Patents patents.google.com/patent/US12548565B2/en → the References Cited and Cited By sections (the page exists; my search snippets confirm the record but not the citation block).
4. Global Dossier (USPTO/EPO) for the 17/867,091 file wrapper, to see which references the examiner actually applied versus merely cited.

---

6. Explicit uncertainties / corrections to the cross-referenced sections

1. No verified (56) citation list for '565. This is the central gap; §1b's 95/8 figures belong to US 11,676,600 B2, not to '565.
2. No prior art verified as anticipating any '565 claim under §102. The only adversarial art I verified (Basye/Shenhav/Parker) is §103 and targets the sibling '750 patent.
3. Do not conflate app. no. 12/548,565 (which is the application number of a different patent — the Samsung-LED-related US 7,846,754) with patent no. 12,548,565. Per your strict rule, these are distinct identifiers and I have kept them separate.
4. Consistency with the earlier sections: my earlier litigation sections reported no direct litigation naming '565 and flagged family-level-only litigation; §2 above is consistent with that — the IPR art targets siblings, not '565. No contradiction arises, but the earlier sections did not cover the (56) citation question at all, which is why this section introduces new material rather than repeating prior content.
5. Date caveat carried forward: several retrieved items (e.g., Unified Patents' portal data) reflect post-issuance updates; the current-date-of-record for this task (2026-04-26) is later than the '565 grant date (2026-02-10), so a live PACER/PatentCenter check is still advisable for anything time-sensitive.

I'll research the prior art available for this patent and its family, since the supplied page record contains only "prior art keywords" and a prior-art date — not an actual list of cited references.

Excellent — I found the key IPR obviousness theory. Let me dig into the specific references Google/Samsung asserted.

Now let me identify the specific references and the ground mappings.

§103 Obviousness Analysis — U.S. Patent No. 12,548,565 B2

0. Source note and a correction to the earlier sections

On "the Prior Art section of this page." The Google Patents record supplied contains no enumerated prior-art list. Its "prior art" content is exactly three fields:

• Prior art keywords: acoustic input, mobile device, acoustic, voice, voice command
• Prior art date: 2013-03-12
• "Family has litigation" banner (family 51529448)

Those are Google's auto-extracted keywords and the priority date — not references. There is no "References Cited," no "Cited By," and no citations table in the record I was given. So there is no page-supplied reference set to combine. Rather than invent one, I have grounded this analysis in (a) the patent's own specification admissions, and (b) the actual prior art that opposing parties have asserted against sibling patents built on the identical 2013-03-12 specification. Both are citable and verifiable, and (b) is the closest thing to an authoritative "prior art section" that exists for this family.

Correction to the two earlier sections — the continuity chain was not unknown; it is stated in the patent itself. Both the summary and litigation sections said the chain into application 17/867,091 "could not be traced." The '565 application's own cross-reference paragraph resolves it:

"This application is a continuation of a continuation of U.S. application Ser. No. 17/397,522 filed Aug. 9, 2021, which is a continuation of U.S. application Ser. No. 15/238,238 filed Aug. 16, 2016, issued as U.S. Pat. No. 11,087,750, which is a continuation of U.S. application Ser. No. 13/795,933 filed Mar. 12, 2013…"
— https://www.patents-review.com/a/20220358924-methods-apparatus-detecting-voice-command.html

Consequences that the earlier sections got wrong or left open:

Why the correction matters for §103: because the '565 is a no-new-matter continuation, its disclosure is byte-for-byte (or near enough) the 2013-03-12 specification shared with the '750 and '461. Any §103 theory that maps onto the '750 or '461 claims maps onto the '565 claims with only the claim-drafting deltas changing. That is the single most important structural fact in this analysis.

---

1. Governing law and the effective filing date

• Application 17/867,091 was filed 2022-07-18, but claims benefit only to applications filed on or before 2013-03-12.
• The AIA first-inventor-to-file provisions apply to applications filed on or after 2013-03-16. Under AIA §3(n)(1), pre-AIA §§102/103 govern unless the application at any time contained a claim with an effective filing date on or after 2013-03-16.
• Working assumption (flag): on the record available, the '565 is a pre-AIA patent. Consequences:
  1. Prior art measured under pre-AIA §102(a)/(b) (publication/patenting before 2013-03-12) and pre-AIA §102(e) (US patents and published applications by their filing date, regardless of publication date).
  2. Pre-AIA §103(c) common-ownership exception is available for §102(e)/(f)/(g) art commonly owned at the time of invention — relevant if any reference is a Nuance/Cerence document.
  3. In an IPR, only patents and printed publications may be used (35 U.S.C. §311(b)). System art — Apple Siri and Samsung S Voice — is therefore unavailable in an IPR but is the centerpiece of Samsung's district-court defense.

Date-sensitivity of the asserted references. This is where most of the fight will be:

Sources: Samsung's contention tables — https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1556712](/patent/1556712)/download-documents?artifactId=ffzqHb1R4bpwnkIaZeQXpDB0E6JkhjbNRGE_UqjZhCtNdwGDrxPWYEg

---

2. The prior-art record actually in play

Group 1 — Samsung's Second Amended Invalidity Contentions (E.D. Tex. 2:23-cv-00482-JRG-RSP), charted against the '461

Primary set: Binder, Gruber, Hwang, Sarin, Hung, plus Kristijansson, Liu, Mozer '171, Newman, Oh, and at least one further reference (US 2015/0106089 A1). Same source as above.

Group 2 — The Google/Samsung IPRs (the only §103 theories ever formally articulated)

IPR2024-01464 (US 11,393,461 — the '565's grandparent):

"The petition alleges that each claim is obvious under 35 U.S.C. §103 when combining the teachings of Sarin, Hwang, Garner, Gruber, and Lebeau, which collectively disclose contextual cues, low-power processing, and server-off-loading. It also argues that discretionary factors under §§ 314(a) and 325(d) favor institution, noting the lack of prior-art consideration during prosecution…"
— https://cases.arcticinvent.com/us/patent/us/petition-for-inter-partes-review-of-us-patent-no-[11393461](/patent/11393461)/ipr2024-01464/

IPR2024-01465 (US 11,087,750 — the '565's parent): parallel petition, supported by the declaration of Mr. Christopher M. Schmandt. Retrieved excerpts show:

• Basye as the two-processor backbone: a low-power DSP (the claimed "first processor") and an "application processing module 112's microprocessor (the claimed 'second processor')" which "require[s] a relatively large power draw" when running on-device ASR.
• Basye–Shenhav–Lebeau mapped to claims 10 and 18.
• Waters (US 9,020,459) cited for the motivation to power-gate/duty-cycle a DSP.
• Sarin (US 9,031,847) cross-referenced.
• Petitioners' own framing: "The Office did not consider a wealth of prior art during prosecution that discloses and/or suggests the alleged invention" and "[t]o the best of Petitioners' knowledge, the '750 Patent… has not been challenged in any other [IPR]" — and "'461… is the parent of the '750 Patent."
  — https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1556713](/patent/1556713)/download-documents?artifactId=HGmfq2QsTTu3s6cZGyw--SD4juERg2Nwf2GacnyDfrJdFTn-bEYjHRc and …artifactId=LHkyR1Ak9Z4VZaksevYZrPCz-rQGO5J0TqN0nirYneS9ujCBLmWRZ0Y

Outcome — critically important: the Board denied institution on 2025-04-23 on discretionary/§314(a) grounds only; Director Review was denied; the Federal Circuit denied the mandamus petition (In re Google LLC, No. 2025-144, order filed 2025-11-06). No tribunal has ever reached the merits of these obviousness theories. — https://cdn.patentlyo.com/media/2025/11/25-144.ORDER_.11-6-2025_2599508.pdf

Group 3 — System art (district court only)

The Federal Circuit order records that Samsung's district-court invalidity contentions "contemplate combining the Siri and S Voice systems with the prior art cited in the petitions." Siri (Oct. 2011) and S Voice (2012) both predate the '565 priority date and are usable as public-use / known-system art under pre-AIA §102(a)/(b) in district court — but not in an IPR.

Group 4 — Corroborating state-of-the-art art (same technical space; date verification required)

Do not conflate Samsung's Hwang (US 2013/0110521 A1, filed 2012-05-30) with Qualcomm's Hwang, Kyu Woong (US 11,205,433; US 12,525,234). Different inventors, different assignees, different documents. Per the operating rule, these are separate identifiers.

---

3. What the claims are, and what a §103 attack therefore looks like

Restating only the structuring point from the prior section: the specification enumerates six independent claim families (A–F) — (A/B) detect a voice command without an explicit trigger; (C/D) plurality of processing stages + contextual cue while in low-power mode; (E/F) two-processor ordering with a disjunctive response trigger. Since I still do not have the granted claim text, the mapping below is to those families, not to claim numbers.

The two strongest formal theories available on this record are exactly the two already formulated against the siblings:

• Theory I (against A–D): Sarin + Hwang + Garner + Gruber + Lebeau — the IPR2024-01464 combination for the '461.
• Theory II (against E–F): Basye + Shenhav + Lebeau (‑ Waters), the IPR2024-01465 combination for the '750.

---

4. Combination-by-combination analysis

Combination 1 — Gruber + Hwang + Sarin (with Garner and/or Lebeau) → claim families A, B

Reference roles (as articulated by Petitioners): the five references "collectively disclose contextual cues, low-power processing, and server-off-loading."

• Gruber (US 2012/0265528 A1) — filed 2011-09-30, published 2012-10-18, i.e. §102(a)/(b) art, five months before the critical date. Consistent with the "Gruber" short name and with the Federal Circuit's reference to Siri being in play, this appears to be Apple's "Using context information to facilitate processing of commands in a virtual assistant" — i.e., the contextual-cue disclosure itself (using device context — location, time, motion, user activity — to condition how spoken input is interpreted). If so, it maps to the claim element "using at least one contextual cue to assist in detecting whether the acoustic input includes a voice command" nearly verbatim, and it is dated safely before the §102(b) bar.
• Hwang (US 2013/0110521 A1) — filed 2012-05-30. §102(e) art only. Supplies the low-power/keyword-detection staging.
• Sarin (US 9,031,847 B2) — filed 2011-11-15. §102(e) art. Supplies the server-offload / distributed-recognition element (relevant if any '565 claim family claims network-offloaded stage performance).
• Garner and Lebeau — fill the remaining stage/architecture limitations. ⚠️ I could not verify the document numbers, titles, or dates of "Garner" or "Lebeau" from the record retrieved. See §9.

Motivation to combine (why a POSITA would have done this before 2013-03-12):

1. The '565 specification itself supplies the motivation. Its Background concedes that (i) voice response systems with manual triggers existed, (ii) explicit voice triggers existed, (iii) both were inconvenient or prohibitive "while operating a vehicle," and (iv) devices routinely slept to save battery. Once the problem ("always-on without draining the battery") is posed as the specification poses it, the solution is the ordinary engineering decomposition: cheap gate first, expensive recognizer second, context to break ties. This is an admission-of-prior-art motivation that is very hard to rebut.
2. Same field, same problem, same components. Gruber, Hwang, and Sarin are all in mobile/handset speech processing. KSR makes the "familiar elements according to known methods" combination obvious; Wyers v. Master Lock permits "logic, judgment, and common sense."
3. Predictable result. Gating an expensive recognizer behind a cheap detector is a power/accuracy tradeoff, not a new mode of operation. The '565 itself frames the invention as "a generally optimal balance" between false positives, false negatives, and battery drain — the textbook KSR "obvious to try" / "predictable variation of a design parameter" posture.
4. Rigid-application risk is low. Gruber's contextual-cue machinery is not bodily incorporated into Hwang's detector; it is used for its known function (conditioning an interpretation decision), which is the classic acceptable combination.
5. No teaching away. Nothing in the asserted art teaches that using non-acoustic context to gate command detection is undesirable.

Predicted defensive counters (Samsung/Cerence and would-be challengers should expect these): (a) Gruber is a disambiguation disclosure, not a detection disclosure — Cerence will argue context there refines an already-received command rather than "assist[ing] in detecting whether the acoustic input includes a voice command"; (b) the §102(e) references require careful proof of filing-date priority; (c) family A's "without requiring receipt of an explicit trigger" is not literally taught by art that relies on keyword spotting.

---

Combination 2 — Basye + Shenhav + Lebeau (+ Waters) → claim families E, F

This is the cleanest structural fit to the '565's two-processor claim families, and it was actually briefed in IPR2024-01465 with claim-level mapping.

Reference roles (per the Schmandt declaration excerpts):

Motivation to combine:

1. Battery life is the explicit design objective, and Basye itself frames the second processor as high-draw but higher-capability — so a POSITA optimizing a low-power-mode voice monitor has an express reason to put the cheap stage on the low-power processor. KSR factor (1): predictable result of using prior-art elements for their known functions.
2. Waters supplies an independent, hardware-level reason to enable/disable processor portions incrementally — the '565's "incremental/progressive wake-up" concept.
3. The disjunctive trigger ("either stage … determines") is arguably the weakest link for a challenger, because a straightforward "cascade" design (stage 1 gates stage 2; only stage 2 concludes) reads on the claim only if the claim's "or" is given its plain disjunctive scope. This is a claim-construction battle, not an art-availability battle — and it is precisely why Petitioners spent claim charts on claims 10 and 18.

---

Combination 3 — Oh, Kristijansson, Newman, Hung, Liu, Mozer '171 → the "contextual cue" and "stage-selection" limitations

This is the secondary/backstop combination for the contextual-cue claims, and it is the one Samsung actually charted:

• Oh (US 2010/0009719 A1; published 2010-01-14) is the strongest reference date-wise: it is clean pre-AIA §102(b) art, immune to the filing-date fights that plague the §102(e)-only references.
• Kristijansson (US 2012/0296655 A1; published 2012-11-22) and Mozer '171 (US 2013/0080171 A1; §102(e) via 2011-09-27 filing) supply motion/context-based handling.
• Newman, Hung, Liu supply further §102(e)-dated architecture and context handling.

Motivation: these references are cited to show that each contextual cue the '565 enumerates (time, location, motion, recent activity, battery state, ambient acoustic profile) was individually known and individually used to condition a device's speech-processing behavior. Once each cue is known and each is known to be usable to gate or bias a decision, combining them for the stated purpose (reduce false activations, save power) is the KSR "known technique, known device, predictable result" case. The '565's own "Design/processing strategy" passage — select stages based on a location cue, favor low-power stages on a low battery — is a direct statement of the incentive.

Weakness: a reference that uses motion to wake the device is not the same as using motion to "assist in detecting whether the acoustic input includes a voice command." If any '565 claim ties the contextual cue tightly to the command determination, the challenger must prove the reference's use of the cue goes to the command decision, not merely to wake-up. Qualcomm US 11,437,031 (hand-over-device → activate ASR) is the archetype of this distinction and would likely be argued both ways.

---

Combination 4 — Qualcomm WO 2013/066523 (alone or as primary) → claim families C, D, E, F

This is the reference I would prioritize verifying, because its disclosure as quoted in the ISR is uncomfortably close to the '565's core:

"transitioning out of a low-power state at a processor … after transitioning out of the low-power state, retrieving audio feature data from a buffer, the audio feature data indicating features of audio data received during the low-power state of the processor"; and "receiving sound data at a first component … performing, at the first component, at least one signal detection operation … and selectively activating a second component … wherein the second component when active consumes more power at the electronic device than the first component when active."
— https://patentimages.storage.googleapis.com/ba/0e/01/0a4ecb92104372/WO2013066523A3.pdf

That is a first-component/second-component, power-tiered, low-power-mode acoustic-monitoring architecture with buffered feature data — i.e., the substance of families C/D and E/F. Its publication date (2013-05-10) postdates the '565 priority date, so it can only be §102(e)/(a)(2) art, and only if its underlying US filing/priority predates 2013-03-12. The WO number implies a 2012 international filing, which would make that plausible but not certain. This must be verified against the US priority document before it is relied upon.

---

Combination 5 — Siri / S Voice system art + the printed publications (district court only)

Per the Federal Circuit: Samsung's contentions "contemplate combining the Siri and S Voice systems with the prior art cited in the petitions." This is the most dangerous art and the least available: system prior art is real, public, and predates the critical date, but it is not usable in an IPR (§311(b)) and requires corroborating documentary evidence and public-accessibility proof in district court. Its natural pairing is: Gruber (the Siri printed publication) as the documentary anchor for the Siri system, plus S Voice as the Samsung commercial implementation, plus the low-power-architecture publications (Basye, Sarin, Hwang, WO 2013/066523).

---

5. Motivation to combine — consolidated KSR rationale

For any of the above combinations, a challenger can assemble the motivation from four largely unassailable sources:

1. Applicant's own admissions. The specification concedes the entire problem statement and the desirability of the goal: manual triggers are "inconvenient, annoying and/or … prohibitive (e.g., when the user is driving)"; conventional systems are "unresponsive when the mobile device is asleep"; the object is to "appear that the mobile device is 'always-on'." Under In re Nomiya / Constant v. Advanced Micro-Devices, admitted prior art and admitted design goals are fair game for motivation.
2. Explicit design tradeoff = predictable variation. The patent's own words: "attempting to generally optimize false positive and false negative rates while keeping power consumption to a minimum"; "the processing strategy … may proceed more conservatively … reserving further processing for when it is necessary." A claim to doing that is a claim to a result, achieved by selecting and ordering known detectors — precisely the KSR "design incentive" and "obvious to try" territory.
3. Same field / known technique improving a similar device. All the cited art is handset/mobile speech processing. 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. No teaching away. The asserted art does not disparage hierarchical detection, contextual gating, or server offload; the record shows the industry moving toward all three (Gruber, Sarin, Basye, WO 2013/066523, and later Qualcomm filings).

---

6. Where the §103 case is likely to be weakest

1. "Without requiring receipt of an explicit trigger" (families A/B). If construed to exclude any wake-word/keword gating, keyword-spotting references cannot satisfy it. Art that detects a question or command directly (the Qualcomm "speaking target detector" concept) does — but that art's date is the problem.
2. The disjunctive "either" trigger (families E/F). Cascade architectures teach stage-1 gating stage-2; the claim's "or" is broader. Expect a construction fight and a written-description fight over whether the 2013-03-12 spec supports it (the '750/'461 spec does recite the disjunctive form, so this is likely defensible).
3. Contextual cue function. Cue-for-wake-up ≠ cue-for-command-detection. Tight claim language here will defeat otherwise good art.
4. §102(e) date fragility. More than half of Samsung's charted set is §102(e)-only. Any failure to prove the reference's filing-date priority (or any pre-AIA §103(c) common-ownership issue, or a §102(e) proviso issue) removes it.
5. Procedural reality. The PTAB has now twice declined to reach these merits on discretionary grounds, and the Federal Circuit refused to compel it. A fresh IPR against the '565 faces the same §325(d)/Fintiv risk — though the '565 is not a parallel-litigation patent, which actually improves the institution calculus relative to its parents.
6. Secondary considerations. The prior section reports a $78.5M E.D. Tex. verdict for Cerence against Samsung (unverified in my search) and Cerence's public statements that the asserted technology is "foundational." If nexus can be shown, that is a real Graham counterweight.

---

7. The elephant: Binder, and the priority/benefit attack

Binder (US 2020/0279576 A1, filed 2020-05-20) appears as the first-listed reference in Samsung's chart against the '461. It cannot be prior art to any claim entitled to a 2013-03-12 effective filing date. Its presence only makes sense if Samsung is asserting that at least some claims in this family are not entitled to the 2013-03-12 benefit date — which would require showing the claim lacks written-description support in the 2013-03-12 specification, sliding the effective date to 2016-08-16 (15/238,238) or 2021-08-09 (17/397,522).

This is the highest-leverage §103 lever on the '565, because:

• A sliding effective filing date unlocks 2014–2021 art — a decade of always-on/wake-word/low-power ASR development (including Qualcomm's two-stage and low-power always-on work and a large body of Google/Apple filings).
• It simultaneously disables the §102(b) grace and §102(e) proviso protections Cerence would otherwise enjoy on the 2013 date.
• A no-new-matter continuation is supposed to be immune to this attack — which is why verifying whether the '565's granted claims contain any new limitation (e.g., a specific contextual-cue formulation, a specific stage-selection rule, or a network-offload recitation) is the first thing to do.

Parallel non-§103 exposure worth flagging: the family's expiration dates diverge (Google Patents shows the '565 at 2034-02-22; the '461 at 2033-03-11). Long chains of continuations of a single 2013 disclosure routinely create obviousness-type double patenting exposure unless terminal disclaimers were filed. That is not a §103 ground, but it is a validity attack that travels with the same record.

---

8. Bottom line

Key caveats to carry forward: (i) the granted claim text of the '565 remains unverified — everything above is mapped to claim families from the specification's summary; (ii) no tribunal has ever reached the merits of any of these obviousness theories — the PTAB twice denied institution on discretionary grounds and the Federal Circuit declined to intervene; (iii) the strongest district-court art (Siri/S Voice system art) is categorically unavailable in an IPR; (iv) the identities and dates of Garner, Lebeau, Basye, and Shenhav could not be verified from the retrieved record.

---

9. Explicit uncertainties and required verification

1. Garner, Lebeau, Basye, Shenhav — unidentified. I have short names and, for Basye and Lebeau, partial teaching descriptions from an expert declaration excerpt. I do not have their document numbers, titles, assignees, or dates. Nothing in this analysis depends on their identity, but a real §103 chart does.
2. Claim text unverified. No granted claim text for US 12,548,565 was available. Independent-claim count, numbering, and dependency are unknown.
3. Gruber identification is an inference. The short name "Gruber," the field, and the Federal Circuit's reference to Siri are consistent with US 2012/0265528 A1 being Apple's "Using context information to facilitate processing of commands in a virtual assistant," but I did not verify the title or assignee. Confirm before citing.
4. Binder anomaly unresolved. Google Patents shows the '565's adjusted expiration as 2034-02-22; patents-review.com shows 2043-06-30 for the same patent. A ~9-year discrepancy between two aggregators, reported verbatim and unreconciled.
5. Date inconsistencies. Several retrieved items (the $78.5M verdict reports, the Amazon ITC institution, 2026 docket activity) are dated May–July 2026, later than the "current date" supplied. Reported as retrieved, not silently reconciled.
6. Pre-AIA assumption. If any '565 claim ever had an effective filing date on/after 2013-03-16, the AIA §102/§103 regime governs and the §102(e) analysis in §1 must be replaced with AIA §102(a)(2)/(d). This changes the treatment of every §102(e)-only row in the table.
7. WO 2013/066523's US priority date unverified. Its value as prior art turns entirely on that date.
8. Prosecution history unknown. The '461's examiner of record (Huy C. Ho) had 91 patent references and 8 NPL items of record. The '565's own record of cited art — and therefore its §325(d) posture — was not retrieved, and the Google Patents page supplied contains no citations table. Obtain the '565 file wrapper before drafting any petition.