An approach for improving repeatability and reliability of non-negative matrix factorization for muscle synergy analysis

The aim of this study was to evaluate non-negative matrix factorization (NMF) and concatenated NMF (CNMF) to analyze and reliably extract muscle synergies. NMF and CNMF were used to extract knee joint muscle synergies from surface EMGs collected during a weight bearing, force matching task. Repeatability and between subject similarity were evaluated for each method using intra-class correlation coefficients (ICCs). High repeatability was found for CNMF (>0.99; 0.99–1.0) compared to NMF (>0.26; range 0.26–0.98). Reasonable consistency across subjects was improved using the CNMF over the NMF approach. CNMF was found to be a more reliable approach than NMF and suitable for between subject comparison of muscle synergies.