DualMyo: Multi-Channel Dual-Stream Transformer Architecture for EMG-to-Digit Classification
Surface electromyography (sEMG) is a robust non-invasive modality for human-machine interaction, yet its application remains largely limited to coarse motor tasks such as grasping or rotation. The decoding of fine motor skills, specifically handwriting, remains a challenging problem with potential relevance for prosthetic control and natural communication interfaces. In this work, we explore a…
DualMyo presents a novel Transformer-based architecture designed for decoding fine motor skills from multi-channel surface electromyography (sEMG) signals, with a specific focus on handwriting. Unlike traditional signal-processing pipelines, DualMyo treats these multi-channel sEMG signals as complex time series, capturing the intricate spatio-temporal dynamics of myoelectric activity. The model employs Patch Embeddings and Rotary Positional Embeddings (RoPE) to enhance its ability to handle these dynamics.
Experiments reveal that DualMyo exhibits strong performance within a single session. Moreover, to tackle the common issues of signal drift and sensor displacement across different sessions, the researchers implemented a lightweight fine-tuning strategy comprising just 10 epochs of training. This approach allows DualMyo to adapt effectively to the variability between sessions, achieving an impressive 91% accuracy with just two examples of each digit.
These findings represent a significant advancement in the development of adaptive sEMG-based handwriting interfaces, offering a promising direction for future research. However, further validation will be necessary to ensure its effectiveness for real-time, multi-subject deployment and neuromuscular control applications.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
