Improveci EMG pattern rocognition using polar segmentation of the feature space

In this paper, different segmentation schemes of the EMG feature space are considered in order to implement a myoelectric prosthetic control with a minimum classification error. The pattern classification algorithm itself is based on heuristic rules. It is shown that the use of polar coordinates instead of cartesian ones may lead to a lower misclassification rate. The computation complexity of this kind of segmentation can be easily handled by current commercially available microprocessors.