Learning Vector Quantisation based recognition of offline handwritten signatures

Biometrics, which refers to the identification of an individual based on his or her physiological or behavioural characteristics, has the capability to reliably distinguish between an authorised person and an imposter. This paper presents a low-cost and high-speed Artificial Neural Network (ANN) based offline recognition system of handwritten signatures that is trained with low-resolution and small-sized scanned signature images. The proposed architecture recognises mixed (both Arabic and English) handwritten signatures based on varying parameters and eliminating redundant hidden layer units that learns the correlation of patterns. Empirical results yield an accuracy rate of 98.7% for an unseen 150 signatures of varying covered areas of 10 persons on the network that is trained with another 120 images. The robustness of the proposed algorithm is demonstrated by calculating standard deviation of 50 classifiers.