Offline Handwritten Multilingual Numeral Recognition Using CNN

In developing automated systems, such as those used for sorting mailed addresses and reading license plates, numerical recognition plays an essential part. Due to the increase in international communication and business transactions, numerical recognition systems that recognize many languages are beneficial in today’s world. This is particularly true in multilingual nations where numerous languages are spoken concurrently. Recognition of printed numbers is far easy than recognizing Numerals written by hand because of the variety and complexity of people’s handwriting. Overall, having poor handwriting may appear to be a small problem, but it can have serious repercussions on both the individual and society. So, it is important to promote legible and clear writing as a vital communication skill. As a result, the creation of a multilingual handwriting system is seen as a topic that is both significant and contentious. To solve this problem, we proposed CNN-based approach to recognize the multilingual Numerals. Numeral recognition is included in the approach that we have proposed. Its goal is to manage the recognition of images, including several language Numerals, within the framework of the suggested system. Fourteen distinct languages were used in the evaluation of the system that was proposed. The proposed system utilizes MNIST-MIX multilingual Numeral dataset. The proposed system achieved promising results.