Analysis of transient states in EMG signals with applications to robotic prostheses

The purpose of this thesis is to analyze and classify EMG signals for prosthetic control. For transradial amputees the electrical signal of muscles is acquired, processed and classified with the final scope of moving a prosthetic hand. This signal has a first transient state related to the dynamic part of the movement starting from rest and a subsequent steady state which corresponds to the sustained and voluntary contraction. Up to now the classification has been performed only on the steady state, instead in this thesis we will include also the transient. The classification of the transient, due to its instability, is more difficult than for a stationary signal, however it can help to speed up the time necessary to move the prosthesis. In order to do that we use different instruments of pattern recognition and we show that in many situations the transient already provides a good classification, compatible with the steady state. Nevertheless the results depend on the abilities of the single patient.