Identification of Motor Units from High-Density Surface Electromyography Signals Using a Long Short-Term Memory Network for Human-Machine Interaction

This paper introduces a novel approach for decoding motor intentions by identifying motor units (MUs) from high-density surface electromyography (HD-sEMG) signals using a Long Short-Term Memory (LSTM) network, focusing specifically on human-machine interaction applications. By optimizing the Latent Component Analysis (LCA) method with LSTM, our framework enhances the accuracy and efficiency of MU identification, addressing critical needs in real-time and responsive human-machine interfaces. The proposed method significantly reduces computational complexity relative to traditional techniques, making it particularly suitable for dynamic and intuitive interaction systems. Experimental results demonstrate that our approach not only achieves high accuracy but also operates with low latency, positioning it as a robust solution for advancing humanmachine interaction technologies.