Comparison of sEMG Encoding Accuracy Across Speech Modes Using Articulatory and Phoneme Features

2026-04-20

AI Summary

This study investigates whether surface electromyography (sEMG) signals can be effectively encoded across overt speech, silent reading, and imagined (silent) speech, and evaluates the suitability of different speech representations as intermediate targets. Using multivariate temporal response function (mTRF) models with elastic net regularization within a sentence-level cross-validation framework, the authors systematically compare the linear predictive performance of Speech Articulatory Features (SPARC) against phoneme one-hot encodings for modeling sEMG envelopes. Results demonstrate that SPARC significantly outperforms phoneme-based representations across nearly all electrodes and speech conditions; overt and silent reading yield comparable performance, while imagined speech remains substantially above chance. The mTRF weights exhibit consistent anatomical interpretability. This work provides the first evidence of SPARC’s robust encoding capability for sEMG across multiple speech modalities, establishing it as a preferred intermediate target for silent speech modeling.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Natural Language Processing: Speech
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We test whether Speech Articulatory Coding (SPARC) features can linearly predict surface electromyography (sEMG) envelopes across aloud, mimed, and subvocal speech in twenty-four subjects. Using elastic-net multivariate temporal response function (mTRF) with sentence-level cross-validation, SPARC yields higher prediction accuracy than phoneme one-hot representations on nearly all electrodes and in all speech modes. Aloud and mimed speech perform comparably, and subvocal speech remains above chance, indicating detectable articulatory activity. Variance partitioning shows a substantial unique contribution from SPARC and a minimal unique contribution from phoneme features. mTRF weight patterns reveal anatomically interpretable relationships between electrode sites and articulatory movements that remain consistent across modes. This study focuses on representation/encoding analysis (not end-to-end decoding) and supports SPARC as a robust and interpretable intermediate target for sEMG-based silent-speech modeling.

Problem

• sEMG
• speech modes
• articulatory features
• phoneme features
• encoding accuracy

Innovation

• SPARC
• sEMG
• mTRF
• articulatory features
• silent speech