Automatic Speech Recognition Patents: Top Companies & Trends 2026
- 53.9% concentration. The top 5 assignees hold 433 of 804 records in scope — over half the field sits with a handful of filers.
- Filing has plateaued, not accelerated. Activity peaked at 109 records in 2023, and 2026 volume (partial year) sits well below that, after flat-to-declining growth from the 2022 midpoint of 90.
- Almost everything sits in one IPC subclass. G10L (speech & audio analysis/synthesis) covers 98.4% of the 804 records, leaving thin, specific side branches in G06N, G09B and A61B.
What the automatic speech recognition patent record shows
Automatic speech recognition patenting sits almost entirely inside G10L, the speech and audio analysis/synthesis subclass, with G06N (AI models) and G06F (digital data processing) as the two consistent secondary classes. That pattern tells a simple story: the core recognition pipeline — acoustic modelling, decoding, noise handling — is where claim density lives, and everything else (telephony integration, transducers, education tools, diagnostic use) is a smaller adjacent layer built on top of it.
The assignee base is top-heavy. Five companies account for 53.9% of the 804 records in scope, and ten account for 66.0%. Beyond that top tier the ranking runs into a long tail of single- and low-filing entrants, which is typical of a field with an established technical core and a wide fringe of applied variants.
Let an AI agent run this analysis on your own technology
Pick a task. Every answer cites the patents behind it.
Filing trend and technology composition
Publication for the most recent year always lags actual filing by roughly 18 months, so the 2026 figure below understates real activity. Read the trend as a plateau from 2022 onward rather than a cliff.
A 2023 peak followed by decline
Filings rose from 35 in 2017 to a peak of 109 in 2023, held near that level through the 2022 midpoint of 90, then fell — down to 9 by 2026 (a partial year). The shape points to a field where the core claim space has been substantially staked out rather than one still in a filing ramp.
One dominant class, several thin ones
G10L covers 98.4% of the 804 records in scope, confirming that nearly every filing touches core speech/audio processing. G06N (19.7%) and G06F (13.4%) are the meaningful secondary layers — AI model architecture and general data processing applied to ASR — while H04M, H04R, G09B, A61B and G06K each account for under 3% and mark narrower, more specific application areas.
Shares are the percentage of the 804 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Automatic Speech Recognition with Eureka
This page is one run against one query. Ask Eureka your own question about automatic speech recognition and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative and most-cited filings
US12087276B1 — Automatic speech recognition word error rate estimation applications, including foreign language detection
A plurality of audio datasets associated with captured audio are provided to a plurality of automatic speech recognition engines, wherein each of the automatic speech recognition engines is configured to recognize speech of a first language. Word error rate estimates that comprise at least one word error rate estimate for each of the plurality of audio datasets are determined from outputs of the plurality of automatic speech recognition engines. From the word error rate estimates, audio in the plurality of audio datasets is determined to include speech in a second language.Filed by Cisco Technology, Inc., granted 2024-09-10 — an example of ASR claim scope extending beyond transcription accuracy into diagnostic uses like language detection via error-rate estimation.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US6615172B1 | Intelligent query engine for processing voice based queries | 812 |
| 2 | US7725307B2 | Query engine for processing voice based queries including semantic decoding | 605 |
| 3 | US10573312B1 | Transcription generation from multiple speech recognition systems | 443 |
| 4 | US20200175961A1 | Training of speech recognition systems | 410 |
| 5 | US20040117189A1 | Query engine for processing voice based queries including semantic decoding | 383 |
| 6 | US6366882B1 | Apparatus for converting speech to text | 371 |
| 7 | US7761296B1 | System and method for rescoring N-best hypotheses of an automatic speech recognition system | 299 |
| 8 | US20140163981A1 | Combining Re-Speaking, Partial Agent Transcription and ASR for Improved Accuracy / Human Guided ASR | 275 |
| 9 | US20200243094A1 | Switching between speech recognition systems | 265 |
| 10 | US20100121638A1 | System and method for automatic speech to text conversion | 194 |
Citation counts favour older filings simply because they have had more time to be cited — treat them as a signal of influence within this searched corpus, not a ranking of current technical importance.
Each row carries its publication number; clicking a row searches Eureka by that number.
Put your own technology through the same analysis
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →What the numbers mean for filing strategy
Three patterns matter more than the raw counts: where claim density sits, how concentrated ownership is, and what the trend line implies about remaining headroom.
Ownership is top-heavy
With the leader alone holding 190 records and the top 5 combining for 433 of 804, new entrants are filing into a space where core acoustic-modelling and decoding claims are already dense. The gap between the leader and fifth place (190 vs 25) shows the concentration is driven by one dominant filer, not an even top tier.
Filing has cooled from its 2023 peak
Recent-year momentum data shows the largest historical filers posting flat or negative year-on-year change, several at or near zero for the latest year. That is consistent with an established field where core claims are staked, not one in early growth.
Claim density sits in one subclass
Nearly every record in scope carries a G10L classification, with G06N and G06F as the two IPC classes that consistently pair with it. Classes below 3% — H04M, H04R, G09B, A61B, G06K — mark where ASR intersects telephony, transducers, education and diagnostics, and where filing volume is thin relative to the core.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to automatic speech recognition, with the prior art for and against each one.
Who is filing, and who has slowed
The ranked leaders capture the field's concentration, but recent-year momentum tells a different story than cumulative counts: several of the largest historical filers show little or no activity in the latest year.
A single dominant filer
The top-ranked assignee holds 190 of 804 records, more than double the fifth-place holder's 25 — a gap that suggests one company built out foundational claim coverage well ahead of the rest of the field.
Cumulative leaders are pulling back
Several of the highest-ranked assignees by total record count show sharp year-on-year declines or zero filings in the latest year, indicating the current wave of activity is not coming from the same names that built the historical base.
Co-filing is limited and concentrated
Only 10 co-assignee pairs appear in the dataset, with the strongest links tied to one large filer's internal inventor teams rather than cross-company joint filings — collaboration here is mostly organisational, not inter-company.
| Assignee | Recent year | YoY |
|---|---|---|
| Google LLC | 3 | -75% |
| Microsoft Technology Licensing, LLC | 0 | -100% |
| International Business Machines Corporation | 0 | — |
| Intel Corporation | 0 | -100% |
| AT&T Intellectual Property II, L.P. | 0 | — |
| Nuance Communications, Inc. | 0 | — |
| Samsung Electronics Co., Ltd. | 0 | — |
| Sorenson IP Holdings, LLC | 0 | -100% |
Where to take this next
The dataset points to specific next checks rather than a single conclusion — concentration at the top, a cooling trend, and thin coverage outside the G10L core all suggest different follow-up work depending on whether you are filing, licensing or monitoring competitors.
Check freedom-to-operate against the top assignees
With 53.9% of records held by five companies, a freedom-to-operate review focused on the leader's claim scope is likely to surface more blocking prior art than a broad field-wide search.
Run an FTO check in EurekaWatch for new entrants in thin IPC branches
Classes under 3% of records — H04R, G09B, A61B, G06K — carry far less claim density than G10L and may offer more room to file defensible claims.
Explore white space in EurekaTrack momentum, not just cumulative rank
Several top-ranked assignees by total count show flat or negative recent-year filing. Monitoring current-year activity separately from historical rank gives a more accurate read on who is actually active now.
Set up assignee monitoring in EurekaCommon questions about ASR patents
The ranked leader holds 190 of the 804 records in scope, well ahead of the fifth-place assignee at 25. The top 5 assignees combined account for 53.9% of all records, and the top 10 account for 66.0%, so ownership in this field is heavily concentrated in a small group of large technology companies rather than spread evenly. Beyond the top tier, the ranking runs into a long tail of companies with only a handful of filings each.
No — filing activity peaked at 109 records in 2023 and has declined since, with the 2026 count (a partial year) sitting at 9. Because publication typically lags actual filing by around 18 months, the most recent year always understates true activity, but the trend from the 2022 midpoint of 90 records onward is flat to declining rather than growing. Several of the historically largest filers also show zero or sharply negative year-on-year filing in the most recent period.
Almost all records — 98.4% of the 804 in scope — carry a G10L classification, covering core speech and audio analysis and synthesis. G06N (AI model architectures) and G06F (digital data processing) are the two consistent secondary classes, appearing in 19.7% and 13.4% of records respectively. Smaller, more specific classes like H04M (telephony), H04R (audio transducers), G09B (education), A61B (diagnosis) and G06K (data recognition) each cover under 3% of records and mark narrower application niches.
The thinnest IPC classes relative to the 804-record total — H04R, G09B, A61B and G06K, each under 3% of records — mark areas where ASR intersects audio hardware, education tools and diagnostic applications but has not been heavily claimed. These branches sit outside the dense G10L core and may offer more room to file defensible claims, particularly around noise-robust decoding, accent and dialect coverage, and domain-specific adaptation, which are named in the underlying search criteria but not saturated by any single large filer.
US12087276B1 is assigned to Cisco Technology, Inc. and was granted on 2024-09-10. It covers a method of running multiple audio datasets through multiple ASR engines, each tuned to a first language, to generate word error rate estimates, and then using those estimates to detect when audio actually contains a second language. It is a representative example of ASR patent scope extending past plain transcription accuracy into diagnostic and language-detection applications, and anyone building similar error-rate-based language detection should review its claims closely.
Research Automatic Speech Recognition in depth with Eureka
Go past this page: query the whole automatic speech recognition corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company’s registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.