Effects on accuracy of Uyghur handwritten signature recognition
In this paper, an approach for off-line Uyghur signature recognition is proposed. The signature images were preprocessed using improved techniques adapted to the Uyghur signature. The preprocessing are included noise reduction, binarization, normalization and thinning. Two types of preprocessing steps were conducted with and without thinning. The directional features, global baseline, upper and lower line features, local central features were extracted respectively after the two kinds of preprocessing. Experiments were performed selecting Euclidean distance and Chi-square distance based measure methods and using K nearest neighbor classifier for Uyghur signature samples from 50 different people with 1000 signatures. A correct recognition rate of 96.0% was achieved with thinning. The experimental results indicated that thinning has significant importance to the extracted features and its effects to the accuracy were related with the nature of extracted features.
|Keywords||K-NN classifier, signature recognition, thinning, Uyghur|
Ubul, K. (Kurban), Adler, A, & Yadikar, N. (Nurbiya). (2012). Effects on accuracy of Uyghur handwritten signature recognition. doi:10.1007/978-3-642-33506-8_67