On automatic identification of upper-limb movements using small-sized training sets of EMG signals

We evaluate the performance of a variety of neural and fuzzy networks for discrimination among three planar arm-pointing movements by means of electromyographic (EMG) signals, when learning is based on small-sized training sets. The aim of this work is to underline the importance that the sparse data problem has in designing pattern classifiers with good generalisation properties. The results indicate that one of the proposed fuzzy networks is more robust than the other classifiers when working with small training sets.