Towards Real-Time Decoding of Motor Unit Firing Events and Resulting Muscle Activation During Human Locomotion and High-Force Contractions

Interfacing with the central nervous system is es-sential for developing personalized neuro-rehabilitation strategies. High-density electromyography (HD-EMG), together with blind source separation (BSS) techniques, enables decoding motor unit (MU) firing activity in a non-invasive manner. However, traditional BSS decomposition techniques are limited to isometric contractions and yield suboptimal results during high-force trials. Additionally, the challenge lies not solely in the decomposition of MUs but also in linking MU firing patterns to their twitch characteristics to decode resultant joint moments accurately. In this work, we introduce a novel, real-time capable neuromuscular framework that enables de-coding accurate firing events and their associated activation profiles simultaneously during both walking and high-force trials. First, we propose a two-stage decomposition to estimate MU filters tailored to walking and high-force HD-EMG data (from the soleus and the tibialis anterior muscles, respectively). Second, we estimate optimal twitch responses that provide accurate MU-specific activation dynamics. For the walking trial, results showed that the estimated activation profiles exhibited a stronger resemblance to the reference moment $(R²=0.89)$ than conventional EMG envelopes $(R²=0.86)$. This suggests that both the decomposition and the estimation of MU twitch properties successfully estimated muscle activation. For the high-force trial (90% of maximum voluntary contraction [MVC]), results showed a broader diversity of decoded MUs with recruitment thresholds ranging from 20% to 70%MVC. Moreover, our algorithms operated within times (< 9 ms) well below the neuromechanical delay. Our study presents for the first time an online-ready method-ology to decode MUs from the soleus muscle during walking with sufficient detail to account for observed ankle moment trends. This has multiple implications for developing neuro-rehabilitation devices that can adapt more effectively to a patient's individual needs and progress.