Analysis and Modelling of Novel Surface Electromyography Signals
This project investigated non-invasive muscle source localisation using high-density surface electromyography (HDsEMG) recorded with the COSIMA prototype, an innovative device with dry electrodes that reduce wiring.
The aim was to develop and validate methods for electrical source mapping, including signal preprocessing, modelling, and analysis.
sEMG signals were recorded during various hand contractions and compared with the open source putEMG dataset. After filtering and segmenting, contraction phases were isolated, and heat maps generated to visualize surface muscle activity consistent with anatomical expectations.
Due to COSIMA’s limited coverage, volume source estimation was performed using only the putEMG dataset. A cylindrical forward model and inverse modelling algorithms were used to localize internal sources, validated through 3D reconstruction and MRI alignment.
Model performance was evaluated using explained variance, Dice coefficients, and confusion matrices. Results showed accurate localisation for flexion and extension, though deeper contractions showed more variability. Inter-subject differences in muscle activation locations were also observed.
Overall, COSIMA shows strong potential as a compact sEMG system for applications in prosthesis control and neuromuscular diagnostics.
Poster du travail de bachelor de Fantin Melmoux
