A Set of Features for Text-Independent Writer Identification

Handwriting is considered to be one of the commonly used biometric modality to identify persons in forensic application. In this paper, we propose a new and efficient approach for off-line text-independent writer identification. Indeed, complete section of connected ink race (Connected-Component or CC), delimited only by where the writer has lifted the pen is inspected in this approach relatively to its aspects of directionality, angularity and length. A set of features is extracted from pre-processed CCs to characterize the writer of a handwritten sample. In order to enhance the performance of our writer identification system, we have combined the features. Experiments are made using handwritings from different people of two datasets, the IFN/ENIT dataset which consists of 1750 documents from 350 writers, and the ICDAR2013 dataset which contains 1000 documents from 250 writers. Our approach showed better properties than most of the surveyed techniques in terms of supported corpus size and identification rate.