Biometric Identification and verification using Deep Learning

Abstract

Because there has not been much research done on the Arabic language which is thought to be more challenging than other languages due to it has multiple cases and forms for a single letter—there is still interest in developing a system to verification the writer\\'s identity by he/she handwritting. These systems are thought to be essential in resolving complex legal issues involving security agencies and other issues.The foundation of our novel approach is the integration of a Harris corner detector with a Convolutional Neural Network (CNN).We also introduce a data augmentation algorithm to enhance the quality of the data. We tackle several problems related to this point by suggesting a method that can identify handwritten Arabic texts without the need to separate the characters,words to lines and by using upsampling to make the handwritten area of the image larger and more readable and this model work in offline mode with value 99.66% on the KHATT dataset.