Writer identification using super paramagnetic clustering and spatio temporal neural network

Abstract

This paper discusses use of Super Paramagnetic Clustering (SPC) and Spatio Temporal Artificial Neuron in on-line writer identification, on Farsi handwriting. In online cases, speed and automation are advantages of one method on others, therefore we used unsupervised and relatively quick clustering method, which in comparison with conventional approaches, give us better result. Moreover, regardless of various parameters that available from acquisition systems, we only consider to displacement of pen tip at determined direction that lead to quick system due to its quick preprocessing and clustering. Also we use a threshold that remove displacement between disconnected point of a word that lead to a better classification result on on-line Farsi writers. © 2009 Springer-Verlag Berlin Heidelberg.