Correlation in time-frequency domain among EMG signals from muscles recruited by the same synergy during walking

This study proposes and validates an advanced method, the pooled scalogram, for quantifying muscle synergies in the time-frequency domain. The pooled scalogram, originally developed by Borzelli et al. at the Polytechnic University of Marche, enables a more detailed analysis of muscle co-activation compared to traditional methods. In this research, the pooled scalogram was applied to surface electromyography (sEMG) signals collected from five lower limb muscles, Tibialis Anterior, Lateral Gastrocnemius, Rectus Femoris, Vastus Lateralis, and Hamstrings, during a motor task. The signals were preprocessed using a 50 Hz notch filter, rectified, and normalized to ensure proper analysis. Subsequently, Non-Negative Matrix Factorization (NMF), a method commonly used to decompose sEMG signals into simpler components, namely muscle synergies, was applied, selecting the optimal number of modules to explain at least 90% of the variance in the sEMG signals. This process yields two fundamental matrices, W and C. The W matrix suggests the muscle composition of a synergy, by representing how much each specific muscle contributes to each synergy. On the other hand, the C matrix reflects such temporal organization; it provides an outline of when each synergy is recruited (or not) on motor task accomplishment. The W and C matrices, taken together, offer a full account of the muscle content of the synergies as well their associated activation timing with respect to movement. The results revealed that the Rectus Femoris (RF) and Vastus Lateralis (VL) were the most frequently co-active muscles during the motor task. To identify the synergy where these muscles were most active, the sum of squares of the weights for these muscles was calculated, and the pooled scalogram was generated, including all muscles and excluding contributions below 0.2. The results indicate that the pooled scalogram is highly effective in characterizing and quantifying muscle synergies, with potential clinical applications for assessing neuromuscular conditions and monitoring motor rehabilitation. By identifying key muscles involved in co-activation, this approach opens new perspectives for designing personalized rehabilitation strategies.