Resumen
The paper analytically reviews the methods and algorithms for solving problems arising in the development of writer identification offline systems. The main applied problems that are solved on the basis of the processing and analysis of handwritten text are considered; the classifications of writer identification systems, as well as the structure of offline systems are given. The paper also reviews handwritten text image databases and shows the tasks to which these databases are aimed. An attempt is made to systematize the algorithms for the preliminary processing of handwritten text images, depending on the task they are tackling at the stage under consideration. Based on the analysis of the most common algorithms of forming the handwritten text feature space, these algorithms are classified into ones for extracting geometric, structural, topological, statistical, and spectral features. A review of the algorithms for selecting each category from the listed features is conducted. Recognition methods used to identify the writer are also reviewed. The results of applying these methods, as well as their benefits and drawbacks are presented.