A Methodological Review of an Offline Writer Identification Framework Utilizing Deep Learning and Handcrafted Approaches

Abstract

Handwriting symbolizes a prevalent form of inquired writing and often gains major interest in legal contexts. Since handwriting is a behavioural trait, no two matured handwritings are identical or can be replicated. Thus, it is a highly effective biometrics approach. This work proposes a thorough examination of techniques for recognizing writers. It aims to present an overview of several datasets, techniques for obtaining attributes and classification algorithms (both handcrafted and deep learning based) for writer recognition. This work contributes valuable insights and support to fresh scholars by concisely presenting several feature extraction methodologies and classification strategies necessary for writer recognition across English, Arabic, Western, and other language scripts. Ultimately, we emphasized the obstacles and untouched findings in the discipline of offline writer recognition. At last, we propose possibilities for future exploration.