LiteEMG-FM: An Efficient and Deployable Foundation Model for Robust EMG Sensing

2026-10-01

AI Summary

This study addresses the challenges of cross-subject generalization in electromyography (EMG) signals and the difficulty of deploying existing models on wearable devices in real time. We propose LiteEMG-FM, a foundation model employing a hybrid CNN-Transformer architecture that learns universal representations through multi-dataset pretraining. Furthermore, we introduce a novel hierarchical wake-up mechanism utilizing a lightweight 1D-CNN to filter non-target activities, combined with multiple inference offloading strategies to achieve comprehensive edge-side optimization that effectively balances latency, power consumption, and memory footprint. Experimental results demonstrate that LiteEMG-FM outperforms state-of-the-art time-series foundation models in zero-calibration cross-user and data-scarce scenarios, thereby validating both its computational efficiency and practical deployability on resource-constrained wearable platforms.

Abstract

Electromyography (EMG) signals vary substantially across individuals, body regions, recording sessions, and sensing hardware, limiting the generalization of models for assistive devices and human-computer interaction. Existing time-series foundation models are also computationally expensive for real-time wearable deployment and often fail to capture EMG-specific time-frequency characteristics. We present LiteEMG-FM, an efficient hybrid CNN-Transformer foundation model for practical EMG sensing. Pretrained on 16 diverse upper- and lower-limb EMG datasets, LiteEMG-FM learns representations that generalize across users and datasets. For resource-constrained deployment, we implement a hierarchical wake-up architecture in which a lightweight, always-on 1D-CNN filters rest and non-target activity and activates LiteEMG-FM only for valid gestures. We evaluate full inference offloading, split inference, and full on-device processing, characterizing their trade-offs in latency, power consumption, and memory footprint. Across diverse evaluation settings, LiteEMG-FM outperforms state-of-the-art time-series foundation models and supervised baselines, particularly under zero-calibration cross-participant and data-scarce conditions. These results demonstrate that LiteEMG-FM is an effective, efficient, and deployable foundation model for EMG applications.

Problem

• Electromyography
• Foundation Model
• Generalization
• Wearable Deployment
• Time-Frequency Characteristics

Innovation

• Foundation Model
• Electromyography (EMG)
• CNN-Transformer Hybrid
• Hierarchical Wake-up Architecture
• Edge Deployment