TinyMyo: a Tiny Foundation Model for Flexible EMG Signal Processing at the Edge

Surface electromyography (EMG) is a non-invasive sensing modality used in biomechanics, rehabilitation, prosthetic control, and human-machine interfaces. However, achieving robust generalization across subjects, recording systems, and acquisition protocols remains challenging. While foundation models (FMs) are gaining traction for EMG, existing approaches remain limited to single downstream tasks and lack deployability. We address these limitations.

We present TinyMyo, a lightweight FM based on a Transformer encoder architecture. The model is pre-trained in a self-supervised manner using masked reconstruction on public datasets. With 3.6 M parameters, TinyMyo is designed to support multiple downstream tasks through minimal task-specific head adaptations.

We demonstrate generalization across hand gesture classification, hand kinematic regression, speech production and recognition, with performance comparable or surpassing the state of the art (SoA), and model size below 5 M parameters. We achieve SoA results compared to previous FM-based works on the NinaPro DB5 (87.98%), UCI-EMG (97.1%), and EPN-612 (96.6%) datasets. We demonstrate the first-time deployment of an EMG FM on an ultra-low power microcontroller (GAP9), with an inference time of 0.785 s, energy of 44.91 mJ and power envelope of 57.18 mW, reduced to 0.496 s and 0.089 s inference times with a smaller (1.9 M) model variant on 5 and 1-s data windows.

TinyMyo demonstrates that compact, self-supervised EMG FM can guarantee generalization across multiple downstream tasks while remaining compatible with low-power edge devices.

TinyMyo is the first EMG FM for ultra-low-power edge-devices, enabling scalable and energy-efficient sensing for motor intent decoding, neuromuscular assessment, and human-machine interaction.