Survey Paper on AI-Based Mathematical Handwritten Formula Recognition Using DL

Abstract

Handwritten formula recognition is a rapidly developing area of computer technology, with people having the greatest ability to identify images and items. However, selecting suitable component sets for handwriting recognition is challenging due to the vast variety of writer's handwriting. Many fields, such as medical imaging, science, and archaeology, rely heavily on handwriting, such as forensic medicine. The rapid growth of handheld devices, virtual textbooks, and specialized communication devices has prompted increased attention to handwriting recognition. To train a deep learning network, a dataset of handwriting is constructed, containing 2200 illustrations of each alphabet and a viewable database. Multidimensional Long Short-Term Memory networks are used in emerging strategies for fine Handwritten Text Classification, but they have limitations, such as removing feature vectors close to those extracted by convolution layers, which are less computationally complex.