Exploration of novel descriptors for online writer identication

Abstract

The problem of writer identification refers to the task of deciding on the authorship of a piece of handwritten document by comparing it against a set of samples saved in a database. Technology has aided in the design of hand held devices like tablets that can capture a person's handwriting. The main constituent of these devices is an electronic pen/stylus that helps in obtaining the dynamic information in handwriting through its tip. Online writer identification systems make use of such textual content for deriving the identity of the writer.The present thesis focuses on proposing novel descriptors for a text-independent online writer identification system. In the first part of our work, we represent the spatio-temporal trace from a document with descriptors derived from a codebook. The derived descriptors take into account, the scores of each of the attributes in a feature vector, that are computed with regards of the proximity to their corresponding values in the assigned codevector of the codebook.