On Combinig Deep and Handcrafted Features for Off-line Writer Identification

Abstract

The purpose of this project is to identify the handwritten texts of the individual. In fact, the text is recognized as a means of verifying the identity of the person and will enable the realization of a biometric system of quick and effective control of handwriting, in order to reduce the risk of fraud in a significant way. In this project, we realized a desktop application that allows the user to recognize the handwriting of the individual. We used the histogram of oriented gradients (Hog) and the co-occurrence histograms of oriented gradients (CoHog) function to extract the parameters before applying identification techniques, which will be explained later. These techniques will give us a result that reflects the process followed by our application. The latter has been tested and validated on a test basis.