Myoelectric signal recognition using genetic programming

This thesis presents a new method of myoelectric signal recognition. Myoelectric signals are electric signals generated by the motion of a person's muscle and can be used as control input for prosthetic hands. It uses genetic programming to create a set of equations capable of recognizing three different myoelectric signals. Three different approaches are presented. The first approach uses genetic programming to create three separate equations. Each equation is capable of recognizing a different pair of the three myoelectric signals. The solution is accomplished by the signal that exactly corresponds to two of the three equations. The second approach creates a single equation capable of distinguishing the three signals. The last approach is a hybrid solution. It uses a simple equation to distinguish 90% of the three signals. It then uses a more complicated equation to distinguish the remaining 10% of the signals