Compressed sensing based on rakeness for surface ElectroMyoGraphy

Surface ElectroMyoGraphy (sEMG) is a fundamental tool in medicine, rehabilitation, and prostethics but also made appearance on the consumer world with devices such as the Thalmic lab's MYO. Current state of the art transfers the whole sEMG signal but encounter problems when this signal has to be transferred wirelessly in real-time. To overcome limitations of the current state of the art we propose compressed sensing (CS) as a technique to reduce the size of sEMG data. This work demonstrates the advantage of using a priori knowledge on the sEMG signal by rakeness-based design of a CS acquisition system. Our CS system was shaped on the general purpose data from Physionet and tested on data acquired for a simple hand movement recognition task. Results show that it is possible to significantly reduce the size of transmitted sEMG data while being able to reconstruct good quality signals and recognize hand movemenents.