TEXT CONTENT DEPENDENT WRITER IDENTIFICATION

Abstract

Text content based personal Identification system is vital in resolving problem of identifying unknown document’s writer using a set of handwritten samples from alleged known writers. Text written on paper document is usually captured as image by scanner or camera for computer processing. The most challenging problem encounter in text image processing is extraction of robust feature vector from a set of inconstant handwritten text images obtained from the same writer at different time. In this work new feature extraction method is engaged to produce active text features for developing an effective personal identification system. The feature formed feature vector which is fed as input data into classification algorithm based on Support Vector Machine (SVM). Experiment was conducted to identify writers of query handwritten texts. Result show satisfactory performance of the proposed system, it was able to identify writers of query handwritten texts.