Abstract
The real-time air-writing multilingual datasets are widely used for several purposes, such as handwriting character or numeral pattern recognition. The air-writing systems are commonly used in operation theatres, online education systems, banking sectors, reservation counters, etc. However, the air-written numeral datasets are less for Devanagari and English languages needed for detecting patterns. Therefore, the present article introduces novel datasets written in the air for Devanagari and English. In addition, this article proposes a systematic novel strategy to collect the air-written multilingual numeral dataset from 100 individuals ranging in 20-40 age groups. The Devanagari and English 0-9 digits were ten times written in the air by every individual resulting in 10,000 images for each language. Thus, 20,000 images were generated and stored in the databases. The proposed dataset is freely available and could be a good resource for pattern recognition research.