A Systematic Review of Multilingual Numeral Recognition Using Machine and Deep Learning Methodology

Abstract

The COVID-19 epidemic has altered how people live and interact with one another and tangible objects. The offline optical character recognition (OCR) system has been affected in this regard. Due to its practical uses, the OCR has long been an intriguing subject for researchers. Numerous research reported on offline and online multilingual numeral recognition. The conventional way of air-writing with sensors is costly and time-consuming. This has led to research in air-writing without using sensors for multilingual numeral recognition. However, air-writing without sensors is an emerging topic that needs to be appropriately addressed efficiently and effectively. Air writing without sensors has many advantages, such as being contactless, cost-effective, having real-time and fast processing, and ease of use. Air-writing without sensors can be effectively used in hospital operation theatres, top-secret agencies, online education, reservation counters, banks, airports, and post offices. Therefore, touchless technology is needed in OCR. The paper thoroughly reviews the recent studies executed for offline and online multilingual numeral recognition. We pay particular attention to available datasets and machine and deep learning models used on various datasets for multilingual number identification. This review analyzed work done in datasets using various segmentation, feature extraction, and classification methods. It also focuses on several classification algorithms used and the accuracy obtained. Finally, the paper also elaborates on the applications and challenges of multilingual numeral recognition. This review will benefit numerous researchers working offline and online in multilingual numeral recognition and understanding the systematic approach.