aiqbench analysis
EMGBlend introduces a self-supervised framework that effectively handles the heterogeneity of public surface electromyography datasets through geometry-aware attention and band-restricted spectral targets. By pretraining on 11 diverse sources, the 109M-parameter model demonstrates improved transferability for gesture recognition and force decoding compared to simple data concatenation or random initialization. While cross-person force estimation remains challenging, this work highlights that explicit mechanism design is superior to naive mixing when combining varied bio-signal recordings.
EMGBlend: Heterogeneity-Aware Self-Supervised Pretraining for Gesture and Force Decoding
arXiv:2609.25582v1 Announce Type: cross Abstract: Public surface electromyography (EMG) datasets vary widely in electrode layout, channel count, frequency support, and size. Simply mixing them for pretraining can misalign channel semantics, introduce spectral targets that some devices cannot observe, and let large or high-channel-count datasets dominate learning. We introduce EMGBlend, a self-supervised framework designed around these differences. It combines shared channel patches with geometry-aware attention, restricts spectral targets to each recording's supported frequency band, and balances exposure across data sources. We pretrain a 109M-parameter model on 11 public EMG sources and evaluate it on gesture recognition, continuous-force regression, and contact classification. EMGBlend consistently outperforms matched random initialization and waveform reconstruction controls. Fixed-budget source controls show that multi-source pretraining improves gesture recognition and remains competitive for force decoding. Ablations confirm that geometry, band-aware targets, and source balancing each contribute to transfer, although cross-person NinaPro force estimation remains difficult. Overall, EMGBlend shows how heterogeneous EMG datasets can be combined through explicit mechanism design rather than simple concatenation. Code is available at https://github.com/tamanano/EMGBlend
Abstract:Public surface electromyography (EMG) datasets vary widely in electrode layout, channel count, frequency support, and size. Simply mixing them for pretraining can misalign channel semantics, introduce spectral targets that some devices cannot observe, and let large or high-channel-count datasets dominate learning. We introduce EMGBlend, a self-supervised framework designed around these differences. It combines shared channel patches with geometry-aware attention, restricts spectral targets to each recording's supported frequency band, and balances exposure across data sources. We pretrain a 109M-parameter model on 11 public EMG sources and evaluate it on gesture recognition, continuous-force regression, and contact classification. EMGBlend consistently outperforms matched random initialization and waveform reconstruction controls. Fixed-budget source controls show that multi-source pretraining improves gesture recognition and remains competitive for force decoding. Ablations confirm that geometry, band-aware targets, and source balancing each contribute to transfer, although cross-person NinaPro force estimation remains difficult. Overall, EMGBlend shows how heterogeneous EMG datasets can be combined through explicit mechanism design rather than simple concatenation. Code is available at this https URL
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2609.25582 [cs.LG]
(or arXiv:2609.25582v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.25582 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 263 words
Related coverage
Exploring Hierarchical Interest Representation For Meta Ads Deep Funnel Optimization
Meta Engineering · Jul 15, 2026
HyBRA-EEG: Hyperbolic Global-Local Brain-Region Aggregation for Self-Supervised EEG Representation Learning
arXiv · Oct 14, 2026
TransfHAR: Self-Supervised Wrist Representations for On-Demand Activity Recognition
arXiv · Aug 19, 2026
What Makes Video World Model Latents Action-Relevant: Prediction over Reconstruction
arXiv · Jun 9, 2026