A technique for optimizing electrode placement for electromyographic control of prostheses

We present a technique that enables optimization of Electromyographic (EMG) electrode placement for grasp recognition. Previous works have shown that sophisticated control techniques for prosthetic devices are becoming available; however the issue of electrode placement has yet to be addressed. By processing a rich field of data, it is possible to determine which of the data sets will allow for greatest accuracy in prosthetic control. Data has been collected and processed from 128 sites on a human forearm while two different grasps were performed. Using two different feature extraction techniques - integral of absolute value and differential absolute value - the difference in means between performing each grasp type has been analyzed. This resulted in several regions around the wrist and the elbow that would be optimal for this particular setup. While the optimization process has been used here for discrimination between two particular grasps, it has the potential to extend to any desired actuation pattern.