From Silent Signals to Natural Language: A Dual-Stage Transformer-LLM Approach

Nithyashree Sivasubramaniam

REVIEW 4 major objections 4 minor 19 references

From Silent Signals to Natural Language: A Dual-Stage Transformer-LLM Approach

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A transformer-based ASR plus LLM post-correction lowers silent-speech word error rate from 36% to 30%.

desk verdict Plausible application of standard components to silent-speech ASR, but the reported 6% WER gain is not verifiable without a validation split, error bars, and the ASR training protocol. read the letter →

arxiv 2509.04507 v1 pith:TZQ66MSG submitted 2025-09-02 cs.CL cs.AI

Nithyashree Sivasubramaniam This is my paper

classification cs.CLcs.AI

keywords silentspeechinterfacesurfaceelectromyographyautomaticrecognitiontransformerlargelanguagemodelpost-correctionworderrorratephonemeambiguity

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Silent speech interfaces turn muscle activity into audible speech, but the synthesized audio is noisy and phonetically ambiguous; the paper's premise is that the next bottleneck is recognizing that audio, not generating it. The proposed fix is a two-stage pipeline: a transformer-based ASR replaces the RNN baseline to capture full-utterance context, and a large language model then rewrites the transcript under conservative constraints to fix grammar and resolve ambiguity. On the Digital Voicing test set, the framework lowers word error rate from 36% to 30%, a 16.6% relative (6% absolute) improvement, with average per-utterance time dropping from 1.42 to 0.78 seconds. If this holds, it means LLM post-correction—already effective in conventional ASR—also transfers to the far noisier silent-speech setting, and that text-level repair is a cheap complement to acoustic modeling.

What carries the argument

The load-bearing mechanism is the two-stage decoding chain: (1) a Transformer encoder-decoder that maps EMG-derived features to log-mel spectrograms and decodes transcripts with beam search, replacing the recurrent DeepSpeech recognizer; (2) a GPT-2 post-processor that rewrites the ASR text under a conservative filter—a 0.7 confidence threshold, minimum-length checks, and domain-relevance constraints—so only high-confidence, meaningful corrections survive. The baseline's DTW/CCA alignment strategy (Audio Target Transfer) and HiFi-GAN vocoder supply the training targets and waveforms; the transformer contributes parallel, utterance-level attention, and the LLM contributes linguistic consisten

What would settle it

Compute bootstrap confidence intervals for WER on the ~100-utterance test set; if the 36%-to-30% gap is within sampling noise, the claim fails. A stronger version: freeze all hyperparameters and the 0.7 threshold on a development set before running the test set; if the gain disappears, it was selection. Comparing against a neural LM rescoring baseline at equal compute would also separate semantic repair from generic fluency.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a dual-stage recognition pipeline improves silent-speech intelligibility beyond what the acoustic transducer alone can deliver. Replacing DeepSpeech with a six-layer Transformer encoder-decoder brings WER from 36% to 32.5%; adding GPT-2 as a post-corrector with a 0.7 confidence threshold and conservative edit filters brings it to 30%. The author attributes the gain to two complementary mechanisms: self-attention lets the acoustic model weight clearer frames more heavily and use utterance-level context, while the LLM repairs residual linguistic errors without over-correcting. The result is presented as the first systematic demonstration of LLM-base

Load-bearing premise

The reported 6-point absolute gain rests on the assumption that WER measured on roughly 100 test utterances is stable and that the transformer/LLM hyperparameters and the 0.7 confidence threshold were not tuned to that test set; the paper reports no validation split or error bars.

Editorial extensions

If this is right

Recognition-stage gains are achievable without retraining the EMG-to-speech transducer; the LLM operates purely on text.
The transformer stage alone reduces average per-utterance time from 1.42 to 0.73 seconds, and the LLM adds only 0.05 seconds, so the accuracy gain is nearly latency-free.
Because the gains come from two independent layers, further acoustic improvements should stack with the same LLM correction to push WER below 30%.
The framework gives SSI systems a natural route toward free-form text refinement—style adaptation, dialectal variation, personalization—beyond grammatical repair.
At 30% WER, the output is still far from fluent conversation; the value here is demonstrating a direction, not solving the intelligibility problem.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same two-stage recipe should transfer to other non-acoustic inputs—ultrasound tongue imaging, lip-movement video, articulometry—since the LLM corrects linguistic errors that are modality-independent; a cross-modal test would confirm this.
The 0.7 confidence threshold is the main control knob the paper leaves unexplored: sweeping it would map a correction-versus-hallucination trade-off and probably reveal a range of achievable WERs.
A stricter experimental design would compare the LLM stage against a neural language-model rescoring baseline at equal compute; if the LLM wins only by a small margin, the gain may be generic fluency rather than semantic understanding.
If the authors report per-speaker WER, the headline 6-point gain may concentrate in a few utterances, which would matter for deciding where to spend future effort.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: the paper bolts a Transformer ASR and a GPT-2 post-correction stage onto an existing EMG-to-speech pipeline and reports a 6% absolute WER gain (36% to 30%) on the Digital Voicing test set. The application is new and the direction is sensible, but the evaluation is too thin to verify the gain: roughly 100 test utterances, no validation split before fixing hyperparameters, no error bars or significance tests, and no description of how the Transformer ASR was trained. The central claim should be treated as an existence proof, not a measured improvement.

What's genuinely useful: LLM-based correction hasn't been applied to silent-speech recognition before, and using Gaddy's public dataset makes the numbers concrete. The paper is clearly written and doesn't oversell the absolute performance—30% WER is still high. The runtime figures (0.78s per utterance) are relevant for assistive applications. The pipeline is a straightforward combination of known components, which is fine if the evaluation supports it.

The soft spots are real. Section 2.5 fixes the LLM confidence threshold at 0.7 and mentions 'conservative constraints' without saying how they were chosen. Section 3 describes transformer hyperparameters as 'optimized' and mentions filtering thresholds for the LLM, but no validation split is reported. On a ~100-utterance test set, a few corrected words can move WER by a point, so the 30% result could be partly in-sample fitting. More seriously, the ASR training protocol is absent: the paper doesn't say what data, loss, or split trained the Transformer recognizer. That omission makes the 32.5% intermediate result unreproducible and the comparison with DeepSpeech hard to interpret. The Limitations section lists disfluency and LLM scope, but doesn't acknowledge these evaluation gaps.

Overall, the paper is a plausible workshop-level contribution. It deserves peer review rather than a desk reject, because the idea is testable and the authors have a real dataset. But the referee should demand a held-out validation, per-utterance error bars or a statistical test, and full ASR training details. If those come back clean, the 6% claim is worth taking seriously.

Referee Report

4 major / 4 minor

Summary. The paper proposes a dual-stage pipeline for silent speech recognition: a Transformer-based ASR module replaces an RNN (DeepSpeech) baseline, and a GPT-2-based LLM post-processing module further corrects transcriptions using confidence filtering and 'conservative constraints.' Experiments on the Digital Voicing dataset (≈100 test utterances) report WER reductions from 36% (DeepSpeech baseline) to 32.5% with the Transformer ASR and to 30% with the LLM correction, a 16.6% relative and 6% absolute improvement. The paper claims this is the first systematic application of LLM-based correction in the SSI domain.

Significance. The problem is relevant and underexplored: silent-speech ASR suffers from phonetic ambiguity, and conventional acoustic-only improvements have limited headroom. If the reported improvement is robust, the paper would provide a simple, modular result with practical value for SSI pipelines. The authors use a publicly available dataset (Digital Voicing), which aids comparison, and the high-level architecture is clearly described. The main weakness is that the central experimental claim rests on a small test set with no validation protocol, error bars, or significance testing, and the ASR training details are incomplete. If the evaluation gaps are addressed, the contribution would be acceptable for a conference/short-paper venue; as written, the 6% absolute gain is not verifiable.

major comments (4)

[Section 3, Table 1] The reported 16.6% relative (6% absolute) WER improvement is computed on a test set of approximately 100 utterances, but no error bars, confidence intervals, or significance tests are reported. At this scale, the 3.5-point gap between the Transformer-only (32.5%) and Transformer+LLM (30%) systems, and the 6-point gap over baseline, could be within sampling variability if a few utterances happen to be easy or hard. The sentence in Section 3 stating that the gain is 'significant' is not supported by any statistical evidence. Please report bootstrap confidence intervals, a paired significance test (e.g., McNemar or matched-pairs WER test), and per-utterance score distributions. Section 4 lists limitations but does not acknowledge this evaluation gap.
[Sections 2.5 and 3] The LLM correction relies on a confidence threshold of 0.7, 'conservative constraints,' and 'filtering thresholds to ensure domain specificity,' while the model hyperparameters are described only as 'optimized.' No validation split is described, and no procedure is given for selecting these thresholds. With a test set of ~100 utterances, selecting thresholds to minimize test WER would make the reported 30% result partly in-sample. The authors must either state that all thresholds and hyperparameters were fixed before seeing the test set, or describe the validation split and threshold-selection procedure used.
[Sections 2.3/2.4] The Transformer ASR training protocol is not specified. The paper does not state what input is fed to the ASR (raw synthesized waveform, predicted mel-spectrogram, or log-mel features from Eq. (2)), what transcript labels are used, what loss is optimized (e.g., CTC versus cross-entropy), how many utterances are used for training and validation, or which speakers are held out. Without these details, the intermediate 32.5% WER result is not reproducible and the comparison to DeepSpeech is difficult to interpret. Please provide a complete experimental setup for the ASR module.
[Section 2.4 vs Section 3] The Transformer architecture is specified inconsistently. Section 2.3 (transduction module) lists 6 layers, 8 heads, model dimension 768, FFN 3072, dropout 0.2; Section 3 (for the proposed model components) lists 8 heads, embedding dimension 512, FFN 2048, dropout 0.1. It is unclear which configuration applies to the ASR module and whether both the transduction and ASR transformers use the same settings. This ambiguity prevents the system from being reconstructed from the paper. Please clarify and unify the architecture specifications.

minor comments (4)

[Section 2.2] The Digital Voicing dataset is credited to 'Gaddy' in the text, but Reference [1] has two authors (Gaddy and Klein). Please attribute appropriately (e.g., 'Gaddy and Klein').
[References] Reference [2] has an incorrect venue abbreviation: 'Trans Intrum Mean 73' should be 'IEEE Transactions on Instrumentation and Measurement' (volume 73). Reference [4] is missing the title and venue; it should include the full citation for the Transformer paper.
[Section 3] The GPU is listed as 'NVIDIA RTX 6000A'—please verify the exact model name (e.g., RTX 6000 Ada Generation). Also, the beam search width of 500 is unusually large; a sentence or two justifying this choice would be helpful.
[General] No code, trained models, or data splits are released. Given the small test set and the ambiguity in training details, releasing code/checkpoints or at least the exact data split would greatly improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the WER comparisons are empirical measurements, not derivations from fitted inputs.

full rationale

The paper's central claim is an empirical comparison of WER across three systems (DeepSpeech baseline, Transformer ASR, Transformer+LLM correction) on the Digital Voicing test set. These are measured outcomes, not quantities derived by construction from the model definitions or from any fitted parameter that is then renamed as a prediction. The confidence threshold of 0.7 and 'optimized hyperparameters' are stated without a validation protocol, which is a reproducibility and statistical-validity concern, but the paper never states that these were tuned on the test set, and no equation or construction makes the reported improvement equivalent to the choice of threshold. The reliance on Gaddy's dataset and baseline is an external benchmark, not a self-citation chain, and there are no self-citations by the author. Section 4 lists limitations but omits the evaluation-protocol gap; that omission is a correctness risk, not evidence of circularity. Under the hard rule requiring a specific reduction (Eq. X = Eq. Y by construction, or a fitted parameter renamed as a prediction), no circular step can be identified.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced; the only additions are standard learned components (transformers, LLM, session embeddings).

free parameters (4)

LLM confidence threshold = 0.7
Used to discard low-probability LLM corrections (Section 2.5); no validation-based search described, likely tuned to maximize WER improvement on the evaluation set.
Beam search width = 500
Reported as decoder beam search width (Section 3); unusually large and no sensitivity analysis; affects ASR output and thus the final WER.
LLM max sequence length = 128 tokens
Set for GPT-2 (Section 3); affects the length of corrections and hence the measured WER.
Session embedding dimension = 32
Learned 32-dimensional embeddings to capture inter-session variability (Section 3); a learned input feature whose dimension is chosen ad hoc.

assumptions (3)

domain assumption Gaddy's pretrained EMG-to-speech pipeline and its DeepSpeech baseline WER of 36% are valid comparison points
The paper inherits the baseline from Ref [1] without re-establishing it in this work's setup (Sections 2.1, 2.3).
domain assumption GPT-2 can correct ASR errors on synthesized silent speech with conservative filtering
The paper relies on prior LLM correction literature without testing whether GPT-2's language priors help or hurt SSI-domain output (Section 2.5).
domain assumption No distribution shift between ASR training and evaluation audio
The paper does not state whether the transformer ASR is trained on synthesized EMG speech, ground-truth speech, or a mix; this is required for the reported test-set WER to be meaningful (Section 3).

how reviews work

0 comments

Cite this review

Pith. "Pith review of From Silent Signals to Natural Language: A Dual-Stage Transformer-LLM Approach." pith.science (2026). https://pith.science/paper/TZQ66MSG

@misc{pith2026250904507,
  author       = {Pith},
  title        = {Pith review of: From Silent Signals to Natural Language: A Dual-Stage Transformer-LLM Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TZQ66MSG}},
  note         = {Machine review of arXiv:2509.04507}
}

read the original abstract

Silent Speech Interfaces (SSIs) have gained attention for their ability to generate intelligible speech from non-acoustic signals. While significant progress has been made in advancing speech generation pipelines, limited work has addressed the recognition and downstream processing of synthesized speech, which often suffers from phonetic ambiguity and noise. To overcome these challenges, we propose an enhanced automatic speech recognition framework that combines a transformer-based acoustic model with a large language model (LLM) for post-processing. The transformer captures full utterance context, while the LLM ensures linguistic consistency. Experimental results show a 16% relative and 6% absolute reduction in word error rate (WER) over a 36% baseline, demonstrating substantial improvements in intelligibility for silent speech interfaces.

Figures

Figures reproduced from arXiv: 2509.04507 by the authors.

**Figure 1.** Overview of the proposed SSI recognition framework. The system integrates transformer-based acoustic modeling with an LLM-based post-processing stage for linguistic refinement. 2. METHODOLOGY 2.1. System Overview Our work builds upon the silent speech pipeline proposed by Gaddy, which maps surface electromyography (sEMG) signals to intelligible acoustic speech. The baseline framework includes signal preprocessing, f… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

19 extracted references · 13 canonical work pages

[1]

Instead of relying on acoustic signals, SSIs exploit non -vocal modalities that capture articulatory and physiological activity underlying speech production

INTRODUCTION Silent Speech Interfaces (SSIs) provide an alternative communication pathway by enabling speech synthesis without vocalization. Instead of relying on acoustic signals, SSIs exploit non -vocal modalities that capture articulatory and physiological activity underlying speech production. A variety of input sources have been investigated, includi...

work page
[2]

System Overview Our work builds upon the silent speech pipeline proposed by Gaddy, which maps surface electromyography (sEMG) signals to intelligible acoustic speech

METHODOLOGY 2.1. System Overview Our work builds upon the silent speech pipeline proposed by Gaddy, which maps surface electromyography (sEMG) signals to intelligible acoustic speech. The baseline framework includes signal preprocessing, feature extraction, EMG-to-acoustic transduction with temporal alignment, and waveform reconstruction using a neural vo...

work page
[3]

For comparability, we follow the same hyperparameter settings as in the baseline Digital Voicing study

EXPERIMENTAL RESULTS All experiments are conducted using the NVIDIA RTX 6000A GPU under the PyTorch framework. For comparability, we follow the same hyperparameter settings as in the baseline Digital Voicing study. In addition, the proposed model components are trained with their respective optimized hyperparameters: transformer encoder - decoder with 8 a...

work page
[4]

LIMITATIONS, FUTURE DIRECTIONS, CONCLUSION This paper proposed a dual -stage pipeline for silent speech recognition, combining a Transformer -based ASR for accurate transcription with an LLM -based module for linguistic refinement. Quantitative evaluation demonstrated a consistent reduction in WER across both stages, confirming the effectiveness of integr...

work page
[5]

Digital Voicing of Silent Speech

D. Gaddy, D. Klein, Digital Voicing of Silent Speech, ArXiv Preprint ArXiv:2010.02960 (2020). https://doi.org/10.48550/arXiv.2010.02960

work page Pith review arXiv doi:10.48550/arxiv.2010.02960 2010
[6]

https://doi.org/10.1109/TIM.2024.3449948

Shan Ullah, Deok -Hwan Kim, Multiaccent EMG -to-Speech Optimized Transduction With PerFL and MAML Adaptations, Trans Intrum Mean 73 (2024) 1 -17. https://doi.org/10.1109/TIM.2024.3449948

arXiv 2024
[7]

Gonzalez -Lopez, Alejandro Gomez -Alanis, Juan M

Jose A. Gonzalez -Lopez, Alejandro Gomez -Alanis, Juan M. Martín Doñas , José L. Pérez -Córdoba, Angel M. Gomez , Silent Speech Interfaces for Speech Restoration: A Review, IEEE Access 8 (2020) 177995 - 178021. https://doi.org/10.1109/ACCESS.2020.3026579

work page arXiv 2020
[8]

Gomez , Lukasz Kaiser , Illia Polosukhin , arXiv:1706.03762v7 (2017)

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones , Aidan N. Gomez , Lukasz Kaiser , Illia Polosukhin , arXiv:1706.03762v7 (2017). https://doi.org/10.48550/arXiv/1706.03762

arXiv doi:10.48550/arxiv/1706.03762 2017

Show all 19 references

[9]

Wong , Lusheng Zhang , Keyan Jin , Qiang Zhang, Tianjiao Li , Jinlong Hou , Lidia S

Tao Fang , DF. Wong , Lusheng Zhang , Keyan Jin , Qiang Zhang, Tianjiao Li , Jinlong Hou , Lidia S. Chao , LLMCL -GEC: Advancing Grammatical Error Correction with LLM -Driven Curriculum Learning, arXiv:2412.12541 (2024). https://doi.org/10.48550/arXiv.2412.12541

2024 doi arXiv
[10]

https://doi.org/10.48550/arXiv.1904.13377

NQ Pham, TS Nguyen, Jan Niehues, Markus Müller, Sebastian Stüker, Alexander Waibel, Very Deep Self-Attention Networks for End-to-End Speech Recognition, arXiv:1904.13377 (2019). https://doi.org/10.48550/arXiv.1904.13377

1904 doi arXiv
[11]

J. Wang, S. Hahm, Speaker -Independent Silent Speech Recognition with Across-Speaker Articulatory Normalizations and Speaker Adaptive Training, in: Interspeech, 2015: pp. 2415-2419

2015
[12]

Schultz, M

T. Schultz, M. Wand, T. Hueber, D.J. Krusienski, C. Herff, J.S. Brumberg, Biosignal -based Sp oken Communication: A Survey, IEEE/ACM Trans Audio Speech Lang Process 25 (2017) 2257 -

2017
[13]

Kimura, M

N. Kimura, M. Kono, J. Rekimoto, SottoVoce: An ultrasound imaging-based silent speech interaction using deep neural networks, in: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, ACM, New York, NY, USA, 2019: pp. 1–11. https://doi.org/10.1145/32906...

2019
[14]

Denby, T

B. Denby, T. Schultz, K. Honda, T. Hueber, J.M. Gilbert, J.S. Brumberg, Silent Speech Interfaces, Speech Commun 52 (2010) 270-287. https://doi.org/10.1016/j.specom.2009.08.002

2010 doi
[15]

https://www.nidcd.nih.gov/health/statistics/quick-statistics-voice- speech-language (accessed December 7, 2024)

National Institute on Deafness and Other Communication Disorders, Quick Statistics About Voice, Speech, Language, National Institutes of Health (2024). https://www.nidcd.nih.gov/health/statistics/quick-statistics-voice- speech-language (accessed December 7, 2024)

2024
[16]

M. Kim, B. Cao, T. Mau, J. Wang, Speaker -Independent Silent Speech Recognition from Flesh -Point Articulatory Movements Using an LSTM Neural Network, IEEE/ACM Trans Audio Speech Lang Process 25 (2017) 2323 –2336. https://doi.org/10.1109/TASLP.2017.2758999

2017
[17]

P. Wu, S. Watanabe, L. Goldstein, A.W. Black, G.K. Anumanchipalli, Deep Speech Synthesis from Articulatory Representations, in: ArXiv Preprint ArXiv:2209.06337, 2022. https://doi.org/10.48550/arXiv.2209.06337

2022 doi arXiv
[19]

Scheck, D

K. Scheck, D. Ivucic, Z. Ren, T. Schultz, Stream -ES: Low - Latency End -to-End Speech Synthesis from Neural Electromyography, in: Speech Communication – 15th ITG Conference, VDE VERLAG GMBH, 2024: pp. 200 –204. https://doi.org/10.30420/45616430

2024 doi
[2271]

https://doi.org/10.1109/TASLP.2017.2752365

2017

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

Instead of relying on acoustic signals, SSIs exploit non -vocal modalities that capture articulatory and physiological activity underlying speech production

INTRODUCTION Silent Speech Interfaces (SSIs) provide an alternative communication pathway by enabling speech synthesis without vocalization. Instead of relying on acoustic signals, SSIs exploit non -vocal modalities that capture articulatory and physiological activity underlying speech production. A variety of input sources have been investigated, includi...

work page

[2] [2]

System Overview Our work builds upon the silent speech pipeline proposed by Gaddy, which maps surface electromyography (sEMG) signals to intelligible acoustic speech

METHODOLOGY 2.1. System Overview Our work builds upon the silent speech pipeline proposed by Gaddy, which maps surface electromyography (sEMG) signals to intelligible acoustic speech. The baseline framework includes signal preprocessing, feature extraction, EMG-to-acoustic transduction with temporal alignment, and waveform reconstruction using a neural vo...

work page

[3] [3]

For comparability, we follow the same hyperparameter settings as in the baseline Digital Voicing study

EXPERIMENTAL RESULTS All experiments are conducted using the NVIDIA RTX 6000A GPU under the PyTorch framework. For comparability, we follow the same hyperparameter settings as in the baseline Digital Voicing study. In addition, the proposed model components are trained with their respective optimized hyperparameters: transformer encoder - decoder with 8 a...

work page

[4] [4]

LIMITATIONS, FUTURE DIRECTIONS, CONCLUSION This paper proposed a dual -stage pipeline for silent speech recognition, combining a Transformer -based ASR for accurate transcription with an LLM -based module for linguistic refinement. Quantitative evaluation demonstrated a consistent reduction in WER across both stages, confirming the effectiveness of integr...

work page

[5] [5]

Digital Voicing of Silent Speech

D. Gaddy, D. Klein, Digital Voicing of Silent Speech, ArXiv Preprint ArXiv:2010.02960 (2020). https://doi.org/10.48550/arXiv.2010.02960

work page Pith review arXiv doi:10.48550/arxiv.2010.02960 2010

[6] [6]

https://doi.org/10.1109/TIM.2024.3449948

Shan Ullah, Deok -Hwan Kim, Multiaccent EMG -to-Speech Optimized Transduction With PerFL and MAML Adaptations, Trans Intrum Mean 73 (2024) 1 -17. https://doi.org/10.1109/TIM.2024.3449948

arXiv 2024

[7] [7]

Gonzalez -Lopez, Alejandro Gomez -Alanis, Juan M

Jose A. Gonzalez -Lopez, Alejandro Gomez -Alanis, Juan M. Martín Doñas , José L. Pérez -Córdoba, Angel M. Gomez , Silent Speech Interfaces for Speech Restoration: A Review, IEEE Access 8 (2020) 177995 - 178021. https://doi.org/10.1109/ACCESS.2020.3026579

work page arXiv 2020

[8] [8]

Gomez , Lukasz Kaiser , Illia Polosukhin , arXiv:1706.03762v7 (2017)

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones , Aidan N. Gomez , Lukasz Kaiser , Illia Polosukhin , arXiv:1706.03762v7 (2017). https://doi.org/10.48550/arXiv/1706.03762

arXiv doi:10.48550/arxiv/1706.03762 2017