Extracting Muscle Synergies Using Constrained Non-Negative Matrix Factorization for Motion Intention in Rehabilitation Robotics

Muscle synergy is considered to be composed of a series of skeletal muscles and controlled by a descending neural structure, which is regarded as a potential approach for predicting the motion intention in rehabilitation robotics. Currently, non-negative matrix factorization (NMF) is the most widely used method to extract muscle synergy, but its solution is non-unique and prone to fall into local optima. To address this issue, a muscle synergies extraction method based on constrained non-negative matrix factorization (CNMF) was proposed and applied to analyze the muscle synergies of stroke survivors. The results showed that CNMF had better stability and robustness in contrast to NMF. Therefore, the CNMF-based muscle synergies extraction method proposed in this paper is of great significance to improve the intelligence of rehabilitation robots.