Modified corner line feature based Uyghur off-line signature verification

Handwritten signature verification has been extensively studied in many languages, but there are a few previous studies have been published on Uyghur handwritten signatures due to lack of research and the challenges. A modified corner line features based off-line signature verification method proposed for Uyghur handwritten signature in this paper. The signature images were pre-processed according to the nature of Uyghur signature. Then corner line features (CLF-16, CLF-32, and CLF-64), modified corner line features from two corners (2MCLF-16, 2MCLF-32, and 2MCLF-48) and four corners (4MCLF-16, 4MCLF-32, and 4MCLF-48) were extracted separately. It was selected K Nearest Neighbour (KNN) classifier and Support Vector Machine (SVM) classifier for verifying Uyghur signatures including 180 genuine signatures, 80 random and skilled forgeries from Uyghur handwritten database. Experiments indicated that the SVM classifier showed higher verification results than KNN classifier with CLF and MCLF, and the 2MCLF demonstrated higher accuracy than 4MCLF with the same number of features. The 4MCLF presented higher verification results than 2MCLF after its feature dimensionality was reduced using Principal Component Analysis (PCA), and the system’s False Acceptance Rate (FAR) and False Rejection Rate (FRR) was further decreased to 7.14% and 7.28% with 4MCLF-48 when 120 genuine samples are trained. Experimental results indicated that this kind of modified corner line features based verification method is suitable for the nature of Uyghur handwritten signature, and it can capture the writing style of Uyghur signature more efficiently.