Compressed sensing study for the sEMG data of SCI survivors

Surface electromyography (sEMG) holds great potential in walking function evaluation. Compressed sensing (CS) leverages the sparsity of signals to decrease the number of samples required. In this study, a sEMG CS algorithm for spinal cord injury (SCI) patients based on regularized orthogonal matching pursuit (ROMP) was introduced. It was used to reconstruct multiple sEMG signals collected from SCI subjects. Its performance was compared with orthogonal matching pursuit (OMP). The impact of diverse measurement matrices on reconstruction accuracy was also evaluated. Results indicates that the combination of ROMP with a binary permuted block diagonal (BPBD) matrix outperforms the conventional OMP algorithm.