# Muscle & brain control: the honest hierarchy *Updated 2026-08-07. Part of the [research series](index.md).* !!! abstract "The short version" Consumer "mind control" is mostly muscle. The only reliable switches you can decode from a consumer EEG headset — blink, jaw clench — are **eye and muscle artifacts leaking into the EEG**, so the honest move is to put the electrode on the muscle, where the same signal is orders of magnitude cleaner and costs ~$50 instead of ~$1,000. And even the best muscle interface handwrites at **20.9 WPM** versus speech at ~150 — so the muscle should carry the *intent to speak*, not the words. "Control your computer with your mind" is the most over-promised sentence in consumer tech. This page is the measured version: what electromyography (EMG — muscle electricity) and electroencephalography (EEG — scalp-recorded brain electricity) can each *actually* deliver for controlling a computer in 2026, and why YazSes bet on the muscle. ## The headline result nobody reads carefully Meta's *Nature* 2025 paper on the sEMG wristband is a genuine landmark: **calibration-free gesture decoding across users** (>90% offline on held-out participants, 9 gestures; ~20.9 words-per-minute handwriting) trained on ~11,000 people ([Kaifosh et al.](#ref-kaifosh)). The band shipped in September 2025 — but bundled with $799 display glasses, with a developer kit that exposes **six fixed gestures, no raw EMG, and no Linux** ([Meta dev FAQ](#ref-metafaq)). The released datasets and checkpoints are non-commercial ([repository](#ref-metarepo)). Read the numbers again, though: **20.9 WPM**. Speech dictates at ~150 WPM ([Ruan et al.](#ref-ruan)). The scientific conclusion for a dictation product is not "EMG will replace typing" — it is: > **EMG's job is the *trigger*, not the text.** The voice carries the > bandwidth; the muscle carries the intent to speak. That's a push-to-talk switch that works with your hands full, under a desk, or when a motor impairment makes holding a key difficult. And a trigger has beautifully forgiving requirements: squeeze-onset detection lands at **35–125 ms** with classical threshold/TKEO methods ([Liu et al.](#ref-onset)) — faster than human reaction time. Even continuous grip *force* regresses at r≈0.97 ([Rahimi et al.](#ref-grip)), which is why "light squeeze = talk, hard squeeze = command" is on our roadmap. ```mermaid flowchart LR A[Forearm electrode\nMyoWare 2.0 ~$50 /\nBioAmp EXG Pill] --> B[Microcontroller\nonset detection 35–125 ms] B -- "YESP over USB serial\nHOLD_START / HOLD_END" --> C[YazSes daemon] C --> D[Same pipeline as the\nhold-to-talk key] D --> E[Squeeze = speak a command\nor dictate hands-free] ``` YazSes speaks this protocol today (`[emg] device_port`); the sensors are open-hardware DIY parts (SparkFun's MyoWare 2.0 muscle sensor at ~$50; the CERN-OHL-licensed BioAmp EXG Pill), and the canonical Python stacks for fancier hardware are [LibEMG](https://libemg.github.io/libemg/) and MIT-licensed [BrainFlow](https://brainflow.org/) — both offline, both Linux-native. ## Why EEG loses this contest (for now) Consumer EEG headsets (Muse, Neurosity Crown, OpenBCI) record real brain signals. The question is what's *decodable* as a reliable command: | Signal | Best measured performance | Fatal problem for daily control | |---|---|---| | Blink (EOG artifact) | 99.5% accuracy, 1.3 s response, **0.10 false positives/min** ([Li et al.](#ref-eogswitch)) | 0.10 FP/min ≈ **~50 phantom activations per workday** | | Jaw clench (EMG artifact on EEG) | ~90%, sub-second possible | Confounded by chewing — and by **talking**, fatal in a dictation tool | | SSVEP (flicker-evoked) | **>12 bits/min** online with a 14-channel consumer headset ([Lin et al.](#ref-ssvep)); far higher only on wet-electrode lab rigs ([Chen et al.](#ref-chen)) | Needs occipital electrodes no headband has, plus permanently flashing on-screen stimuli | | P300 | 20–70 bits/min | Flashing grids + per-session calibration | | Motor imagery ("think left hand") | 70–85% *binary* after 3–10 training sessions | **BCI "inefficiency" — some users never produce usable patterns at all — is a known, unsolved problem** ([Tibrewal et al.](#ref-mi)); nobody accepts a talk key that simply never works for part of its audience | Note the pattern: the only *reliable* "EEG" switches — blink, jaw clench — are not brain signals at all. They are **muscle and eye artifacts** leaking into the EEG. At which point the honest engineering move is to put a proper electrode *on the muscle*, where the same signal is orders of magnitude cleaner: ```mermaid flowchart TD Q{You want a hands-free\nswitch signal} --> A[EEG headband\n$250–1,000] Q --> B[EMG electrode\n~$50 DIY] A --> C[Decodes muscle/eye artifacts\nthrough the skull, noisily] B --> D[Measures the same muscle\ndirectly at the source] C --> E[~50 phantom triggers/day\nbest case] D --> F[35–125 ms onset,\nfalse-activation rate you can tune] ``` Worth noting what *does* fix the false-positive problem in the EEG literature: switching signal entirely. A breath-hold switch read from respiration-modulated photoplethysmography reached **0.02 false operations/min**, five times better than the blink switch ([Han et al.](#ref-ppg)). The lesson generalises — pick the physiological channel where your intended action is loudest, rather than trying to filter it out of a noisier one. This is the reasoning recorded in ADR-v2-129: **consumer-EEG triggers are rejected** — not because BCIs aren't coming, but because in 2026 a dedicated EMG channel strictly dominates on signal-to-noise, latency, and false activations. (For completeness: invasive BCIs are a different world entirely, and none of this applies to them.) One metric we can contribute back: Meta's *Nature* paper reports **no false-activation rate** — the single number that decides whether a trigger is livable. YazSes's encrypted learning corpus records every activation locally, so an EMG user can *measure* their own FP rate without sending anything anywhere. We'd love to publish community-collected numbers. ## The accessibility stakes This isn't gadgeteering. The commercial assistive-tech landscape for people who can't use a keyboard is brutal: eye-gaze AAC devices cost **$10,000–20,000** and are Windows/iPad-locked (TD Pilot ~$10k on iPad, TD I-Series ~$20k on Windows; [reporting on access barriers](#ref-kff)), consumer Dragon is discontinued, and the Linux hands-free stack is a graveyard — eViacam unmaintained since ~2019, Google's Project Gameface archived in 2025, dwell-click broken under Wayland. A free, offline, maintained stack that combines voice + gaze dwell + a cheap muscle switch has essentially **no living competitor** on Linux. That is the long-term program these three research pages add up to. ```mermaid flowchart TD S{What can you reliably move?} --> V[Voice] S --> H[A hand, finger or
any single muscle] S --> E[Eyes only] V --> V1["Dictation + spoken commands
~150 WPM · works today"] H --> H1["EMG squeeze as push-to-talk
~$50 DIY sensor · works today"] E --> E1["Gaze dwell as the trigger
research question 4 below"] V1 --> OUT[Text into any application] H1 --> OUT E1 --> OUT ``` ## Silent speech: when the muscle *does* carry the words The claim above — muscle for the trigger, voice for the text — was clean while the only muscle-to-text result was slow handwriting. Between 2020 and 2026 it stopped being the whole story, and the honest version is more interesting. **Gaddy and Klein** were the first to train a model on EMG recorded during *silently articulated* speech rather than vocalised speech, by transferring audio targets from vocalised to silent signals. Transcription word error rate fell from **64% to 4%** in one data condition and from **88% to 68%** in another ([2020](#ref-gaddy20), improved in [2021](#ref-gaddy21)). Quote the second pair, not the first: 4% is a closed, single-speaker condition, and 68% WER is what open silent speech actually cost at that point. **emg2speech** went further and skipped text entirely. Self-supervised speech (S3) representations turn out to be strongly linearly related to the electrical power of muscle activity — **r = 0.85** — so EMG can be mapped *into* the S3 representation space and synthesised to audio with no explicit articulatory model and no vocoder training. It was demonstrated with a participant with **ALS**, converting orofacial EMG recorded while she silently articulated ([Gowda et al.](#ref-emg2speech)). **emg2qwerty** supplies the missing ingredient — scale. 1,135 sessions, 108 users, **346 hours** of wrist sEMG recorded during touch typing, with baselines built from standard ASR machinery ([Sivakumar et al.](#ref-emg2qwerty)). Its stated central difficulty is not accuracy but **domain shift across users and sessions**, which is the same wall every EMG interface hits. Three 2025–2026 wearables show where the hardware is: | System | Result | Constraint | |---|---|---| | [SilentWear](#ref-silentwear) — 14-ch neck EMG, GAP9 | **20.5 mW**, >27 h on 150 mAh, 15k-param CNN on-device | 8 commands, 4 subjects | | [Textile EMG in headphones](#ref-tang25) — graphene electrodes, ESP32-S3 | **96%** accuracy | 10 control words | | [Soft active EMG interface](#ref-kurotaki) | **97.2% ± 1.3** | 30 words, 3 subjects | The pattern is consistent and it is the single most useful thing on this page: **small closed vocabularies are solved; open dictation from muscle is not.** ### The measured hierarchy, end to end Every number below is from a peer-reviewed measurement, not a demo video. | Path | Rate / accuracy | Source | |---|---|---| | Natural speech | ~150 WPM | [Ruan et al.](#ref-ruan) | | Intracortical speech BCI | **62 WPM**; 9.1% WER on a 50-word vocabulary, 23.8% on 125k | [Willett et al. 2023](#ref-willett) | | Intracortical, rapidly calibrating | **2.5% WER** on a 125k vocabulary | [Card et al. 2024](#ref-card) | | sEMG handwriting (wristband) | 20.9 WPM | [Kaifosh et al.](#ref-kaifosh) | | sEMG silent speech, closed vocabulary | 96–97% on 10–30 words | [Tang](#ref-tang25), [Kurotaki](#ref-kurotaki) | | sEMG silent speech, open vocabulary | ~68% WER | [Gaddy and Klein](#ref-gaddy20) | | Subvocal wearable (first of its kind) | 92%, ~0.5 s latency | [Kapur et al. 2018](#ref-alterego) | The gap between rows five and six is the whole design problem. It is also why the honest reading of "mind-controlled dictation" in 2026 is: **the fast, accurate, open-vocabulary paths all still require surgery** ([Card et al. 2026](#ref-card26) report long-term independent home use), and every non-invasive path buys reliability by shrinking the vocabulary. ### What this changes for YazSes Not the thesis — the *seam*. Splitting the earlier claim in two: 1. **A closed vocabulary of ≤30 silent commands is buildable today.** That maps onto the command grammar YazSes already has, which separates "type this" from "do this" — not onto the dictation path. Silently mouthing *undo*, *new line*, *send* while speech carries the prose is a hybrid the literature supports and nobody ships. 2. **Open dictation stays with the voice**, until a non-invasive path beats 68% WER outside a single-speaker condition. Three concrete consequences we are treating as design work rather than speculation: - **The activation seam is currently trigger-only.** `EMGBackend` implements `HotkeyBackend` — it can say *start* and *stop*, and nothing else. A decoder that produces an intent or a word cannot express it. Widening that protocol to admit an intent-carrying source is the smallest change that makes every system in the table above pluggable. - **The shared representation is the cheap integration point.** If S3 features linearly track EMG power ([r = 0.85](#ref-emg2speech)), an EMG decoder that targets the same latent space a speech encoder already produces can reuse the downstream stack — language model, vocabulary priming, post-processing — instead of rebuilding it. - **Cross-user domain shift is a personalisation problem we already have.** [emg2qwerty](#ref-emg2qwerty) names it as the central obstacle; YazSes already carries per-user calibration and an on-device learning corpus for exactly this failure in the voice path. The machinery is shared even though the signal is not. Robustness has a known shape too: fusing sEMG with lipreading under **cross-modal masking** keeps a system usable when one modality drops out ([del Blanco et al.](#ref-crossmodal)) — the same argument for treating YazSes's several activation sources as a degradation ladder rather than alternatives. And the ceiling is not fixed. Silent-speech decoding from **around-ear EEG** improves substantially with training-set scale ([Inoue et al. 2026](#ref-earEEG)), and a single model can now span **heterogeneous electrode configurations** ([Inoue et al. 2025](#ref-hetero)) — which is the sensor-side equivalent of the device-neutral contract this project uses to keep implementations honest. For a full survey of the sensing modalities, see [Tang et al.'s review](#ref-review). ## Open questions **[Discuss →](https://github.com/MSKazemi/yazses/discussions)** 1. **Measured false-activation rates.** If you run the EMG backend (or build the MyoWare/EXG-Pill rig), what FP/hour do you see across a real workday? This is the number the field doesn't publish — and the one that decides whether a trigger is livable. 2. **Graded squeeze semantics.** Is two force levels (talk / command) the right ceiling, or is a third ("verbatim mode"?) learnable without error? Grip-force regression says the signal is there; the human factors are unmeasured. 3. **Which buyable armband should we adapt next?** MindRove exposes raw 8-channel EMG with an official Linux SDK; second-hand Myo works via `pyomyo`; Mudra Link exposes continuous pressure but lacks a Linux host. What do you actually own? 4. **Dwell-to-talk.** Gaze dwell on a screen zone as the mic trigger (the Look-to-Speak pattern) would close the loop for users who can neither press nor squeeze. What dwell time balances speed against the Midas-touch problem on a webcam-grade signal? 5. **Where is the silent-command ceiling in practice?** The literature reports 96–97% on 10–30 words ([Tang](#ref-tang25), [Kurotaki](#ref-kurotaki)) with 3–4 subjects. Nobody reports what happens across a workday, across sessions, with electrodes re-seated — which is exactly the domain shift [emg2qwerty](#ref-emg2qwerty) identifies as the hard part. A silent command set is only useful if it survives taking the device off and putting it back on. 6. **Should the activation seam carry intent, not just onset?** Today a source can say *start* and *stop*. Widening it so a decoder can deliver a word or an intent would make every system in the table above pluggable — but it also invites a source to inject text the user never reviewed. What is the right confirmation model for a channel with a 3% error rate? !!! tip "If you build the rig, we'll help" A MyoWare 2.0 or BioAmp EXG Pill plus any microcontroller that can speak the YESP serial protocol is enough to drive YazSes today (`[emg] device_port`). Post your build and your false-activation numbers in a Discussion — hardware reports are the contribution this page most needs. ## References **Evidence grade**: *measured* (peer-reviewed measurement), *vendor* (claimed by the maker), *secondary* (review or reporting). 1. Kaifosh, P., Reardon, T. R. et al. "A generic non-invasive neuromotor interface for human-computer interaction." *Nature*, 2025. [doi:10.1038/s41586-025-09255-w](https://doi.org/10.1038/s41586-025-09255-w) — *measured* (20.9 WPM handwriting; >90% offline gesture decoding on held-out users; ~11,000 participants) 2. Meta. "Wearables developer FAQ" (sEMG wristband developer kit scope). [developers.meta.com](https://developers.meta.com/wearables/faq/) — *vendor* 3. Meta Research. `generic-neuromotor-interface` (datasets and checkpoints, non-commercial licence). [GitHub](https://github.com/facebookresearch/generic-neuromotor-interface) — *vendor* 4. Ruan, S., Wobbrock, J. O., Liou, K., Ng, A., Landay, J. "Comparing speech and keyboard text entry for short messages in two languages on touchscreen phones." *arXiv:1608.07323*, 2016. [arXiv](https://arxiv.org/abs/1608.07323) — *measured* (153 WPM speech) 5. Liu, J., Ying, D., Rymer, W. Z., Zhou, P. "Robust muscle activity onset detection using an unsupervised electromyogram learning framework." *PLOS ONE*, 2015. [doi:10.1371/journal.pone.0127990](https://doi.org/10.1371/journal.pone.0127990) — *measured* 6. Rahimi, F. et al. "Simultaneous control of human hand joint positions and grip force via HD-EMG and deep learning." *arXiv:2410.23986*, 2024. [arXiv](https://arxiv.org/abs/2410.23986) — *measured* 7. Li, Y., He, S., Huang, Q., Gu, Z., Yu, Z. L. "A EOG-based switch and its application for start/stop control of a wheelchair." *Neurocomputing* 275, 2018. [doi:10.1016/j.neucom.2017.09.085](https://doi.org/10.1016/j.neucom.2017.09.085) — *measured* (99.5% accuracy, 1.3 s, 0.10 false positives/min) 8. Han, C.-H. et al. "Development of a brain-computer interface toggle switch with low false-positive rate using respiration-modulated photoplethysmography." *Sensors* 20(2), 2020. [PMC7013717](https://pmc.ncbi.nlm.nih.gov/articles/PMC7013717/) — *measured* (0.02 false operations/min) 9. Lin, Y.-P., Wang, Y., Jung, T.-P. "Assessing the feasibility of online SSVEP decoding in human walking using a consumer EEG headset." *Journal of NeuroEngineering and Rehabilitation* 11:119, 2014. [doi:10.1186/1743-0003-11-119](https://doi.org/10.1186/1743-0003-11-119) — *measured* (ITR above 12 bits/min) 10. Chen, X., Wang, Y., Nakanishi, M., Gao, X., Jung, T.-P., Gao, S. "High-speed spelling with a noninvasive brain-computer interface." *PNAS* 112(44), 2015. [doi:10.1073/pnas.1508080112](https://doi.org/10.1073/pnas.1508080112) — *measured* (wet-electrode laboratory rig) 11. Tibrewal, N., Leeuwis, N., Alimardani, M. "Classification of motor imagery EEG using deep learning increases performance in inefficient BCI users." *PLOS ONE*, 2022. [doi:10.1371/journal.pone.0268880](https://doi.org/10.1371/journal.pone.0268880) — *measured* 12. KFF Health News. Reporting on access barriers to ALS communication tools. [kffhealthnews.org](https://kffhealthnews.org/news/medicare-changes-could-limit-patient-access-to-als-communication-tools/) — *secondary* 13. Gaddy, D., Klein, D. "Digital Voicing of Silent Speech." *arXiv:2010.02960*, 2020 (EMNLP). [arXiv](https://arxiv.org/abs/2010.02960) — *measured* (silent-EMG WER 64%→4% in one condition, 88%→68% in another; first training on silently articulated EMG) 14. Gaddy, D., Klein, D. "An Improved Model for Voicing Silent Speech." *arXiv:2106.01933*, 2021. [arXiv](https://arxiv.org/abs/2106.01933) — *measured* 15. Gowda, H. T., Comstock, D. C., Miller, L. M. "emg2speech: Synthesizing speech from electromyography using self-supervised speech models." *arXiv:2510.23969*, 2025. [arXiv](https://arxiv.org/abs/2510.23969) — *measured* (S3 representations predict EMG power at r = 0.85; end-to-end EMG→audio demonstrated with an ALS participant) 16. Sivakumar, V., Seely, J., Du, A., Bittner, S. R., Berenzweig, A. et al. "emg2qwerty: A Large Dataset with Baselines for Touch Typing using Surface Electromyography." *arXiv:2410.20081*, 2024. [arXiv](https://arxiv.org/abs/2410.20081) — *measured* (1,135 sessions, 108 users, 346 h; cross-user and cross-session domain shift as the central problem) 17. Spacone, G., Frey, S., Pollo, G., Burrello, A., Jahier Pagliari, D., Kartsch, V., Cossettini, A., Benini, L. "SilentWear: an Ultra-Low Power Wearable System for EMG-based Silent Speech Recognition." *arXiv:2603.02847*, 2026. [arXiv](https://arxiv.org/abs/2603.02847) — *measured* (14 EMG channels, 20.5 mW, >27 h on 150 mAh, 15k-parameter CNN on-device; 8 commands, 4 subjects) 18. Tang, C., Mallah, J., Kazieczko, D., Yi, W., Kandukuri, T. R. et al. "Wireless Silent Speech Interface Using Multi-Channel Textile EMG Sensors Integrated into Headphones." *arXiv:2504.13921*, 2025. [arXiv](https://arxiv.org/abs/2504.13921) — *measured* (96% on 10 control words) 19. Kurotaki, Y., Yamakoshi, S., Yoshida, R., Isoda, Y., Takano, T., Isano, Y. et al. "Soft Active Electromyography Interface for Machine Learning-Enabled Silent Speech Recognition." *Advanced Intelligent Systems*, 2026. [doi:10.1002/aisy.70440](https://doi.org/10.1002/aisy.70440) — *measured* (97.2% ± 1.3 on a 30-word vocabulary, 3 subjects) 20. Willett, F. R., Kunz, E. M., Fan, C. et al. "A high-performance speech neuroprosthesis." *Nature*, 2023. [doi:10.1038/s41586-023-06377-x](https://doi.org/10.1038/s41586-023-06377-x) — *measured* (62 WPM; 9.1% WER on a 50-word vocabulary, 23.8% on 125k) 21. Card, N. S., Wairagkar, M., Iacobacci, C. et al. "An Accurate and Rapidly Calibrating Speech Neuroprosthesis." *New England Journal of Medicine*, 2024. [doi:10.1056/NEJMoa2314132](https://doi.org/10.1056/NEJMoa2314132) — *measured* (2.5% WER on a 125,000-word vocabulary) 22. Card, N. S., Singer-Clark, T., Peracha, H. et al. "Long-term independent use of an intracortical brain-computer interface for speech and cursor control." *Nature Medicine*, 2026. [doi:10.1038/s41591-026-04414-6](https://doi.org/10.1038/s41591-026-04414-6) — *measured* 23. Kapur, A., Kapur, S., Maes, P. "AlterEgo: A Personalized Wearable Silent Speech Interface." *23rd International Conference on Intelligent User Interfaces (IUI)*, 2018. [doi:10.1145/3172944.3172977](https://doi.org/10.1145/3172944.3172977) — *measured* (92% accuracy, ~0.5 s latency, 10 subjects) 24. del Blanco, E., Gimeno-Gómez, D., Navas, E., Martínez-Hinarejos, C.-D., Hernáez, I. "Cross-Modal Masking for Robust Silent Speech Synthesis Using sEMG and Lipreading." *arXiv:2606.09667*, 2026. [arXiv](https://arxiv.org/abs/2606.09667) — *measured* 25. Inoue, M., Hatakeyama, E., Kita, Y., Sasai, S. "Large-scale training data enhances silent speech decoding with around-ear EEG." *Journal of Neural Engineering*, 2026. [doi:10.1088/1741-2552/ae54d0](https://doi.org/10.1088/1741-2552/ae54d0) — *measured* 26. Inoue, M., Sato, M., Tomeoka, K., Nah, N., Hatakeyama, E. et al. "A Silent Speech Decoding System from EEG and EMG with Heterogenous Electrode Configurations." *arXiv:2506.13835*, 2025 (Interspeech 2025). [arXiv](https://arxiv.org/abs/2506.13835) — *measured* 27. Tang, C., Qi, L., Gao, S., Zhang, Z., Yi, W., Xu, M. et al. "Sensing technologies for silent speech interfaces." *Nature Sensors*, 2026. [doi:10.1038/s44460-025-00010-2](https://doi.org/10.1038/s44460-025-00010-2) — *secondary* (review) 28. Tang, C., Xu, M., Yi, W., Zhang, Z., Occhipinti, E., Dong, C. et al. "Ultrasensitive textile strain sensors redefine wearable silent speech interfaces with high machine learning efficiency." *npj Flexible Electronics*, 2024. [doi:10.1038/s41528-024-00315-1](https://doi.org/10.1038/s41528-024-00315-1) — *measured* *See also: [eye control](eye-control.md) for the gaze half of the hands-free stack, the [accessibility use-case guide](../use-cases/accessibility-rsi-hands-free.md), and the [research index](index.md) for how these channels compare on bandwidth.*