Real-time air-writing numeral datasets are extensively utilized for various applications, including pattern identification of handwritten characters or numerals. The banking industry, reservation counters, online education platforms, operating rooms, and other settings frequently employ air-writing technologies. The English language require more air-written isolated numeric datasets in order to be used for pattern recognition. Thus, for English, the current work presents fresh datasets written in the air. Furthermore, a comprehensive unique technique for air-written real-time English isolated numeral and script categorization recognition is proposed in this paper. We have achieved highest accuracy for English, which is 99.25%. We collected samples from 100 people in age groups spanning from 20 to 40 for each language in order to construct the dataset. Every person wrote the numbers 0 through 9 ten times in the air for each language, producing 10,000 pictures for English language. The suggested dataset is openly accessible and can be a useful tool for study on pattern recognition.