Robust compressive sensing algorithm for wireless surface electromyography applications

Surface EMG (sEMG) can be processed to detect medical abnormalities, activation level, or recruitment order or to analyze the biomechanics of human or animal movement. Today's sEMG systems suffer of limited processing time, limited storage capacity, and high power consumption. The main motivation of this work is to present a new algorithm based on analog-Compressed Sensing (CS) for the receiver side of an ultra-low-power wearable and wireless sEMG sensor. The novel algorithm based on analog-CS at the sensing step attempts to keep the percentage root-mean-square difference (PRD) and compression ratio (CR) in linear relationship that is very important for sEMG data compression to prevent dramatic information loss in high CR situations. The proposed algorithm allows reducing the power consumption to 63%, the PRD to 0.098%, the root mean square error (RMSE) to 0.7%, and signal difference to noise ratio (SDNR) to 0.015%. In addition, the proposed algorithm achieves a good level of accuracy at 98.85% for the reconstruction process.