WRITER INDENTIFICATION BASED ON HYPER SAUSAGE NEURON
Samsuryadi Sahmin, Siti Mariyam Shamsuddin · 2011
This paper proposes biomimetic pattern recognition (BPR) based on hyper sausage neuron (HSN) and applies it in writer identification. HSN is used to cover the training set. HSN's coverage can be seen as a topological product of a one-dimensional line segment and an n-dimensional supersphere. The feature extraction is moment invariants such as united moment invariants (UMI) and aspect united moment invariants (AUMI). The experiments result show that AUMI-HSN method is more effective than UMI-HSN method for identifying the authorship of handwriting.