Static signature verification employing a Kosko-neuro-fuzzy approach

To overcome difficulties in transferring "classical" handwriting examination methods into computer algorithms, a hybrid neuronal system, proposed by B. Kosko (1997), was employed to derive rules for signature region matching. The segmentation of signatures, written on paper documents, into regions is presented and two-stage fuzzy rule learning for finding and tuning the fuzzy rules is discussed. By using Kosko's neuro-fuzzy approach, a region-matching performance of 98% was achieved.