Abstract—Documents of the ancient time gift several opposition for standalone gesture recognition systems, among them, the division and classification steps. Fastidiously gloss wordings square measure is required to coach a system. In some eventualities, written document square measure solely offered at the subdivision level. During this activity, we have a tendency to demonstrate the way to train the system with few tagged information. We have a tendency to additionally propose a model-based social control theme that considers the variability within the writing scale at the popularity section. We have a tendency to apply this approach to the publically offered browse dataset. Our system achieved the competitor result. Humans have distinctive handwriting designs that prove to be an obstacle for handwriting recognition algorithms. To date, multiple researches are done to acknowledge these totally different handwriting designs, most notable mistreatment the synthetic neural network (ANN) with back propagation algorithms that has additionally been verified to relinquish adequately high accuracies. By mistreatment real time method image capturing, this technique and algorithmic rule will be enforced to use multiple written entry information for faculties and universities, wherever the written information of a regular score sheet from totally different people will be transferred to a computer program.