Recognition of Handwritten Mathematical Symbols Affected by Noise using DNN

In all world's countries, mathematical symbols (MS) are international and are hand-written or typed using the same symbol. This is not the case with the alphabet where each country has his own letter. The expression for an equation is written as a combination of letters and MS, e.g. “a+b”. In the case of handwritten documents, proper recognition of the MS is necessary to check if the mathematical expression is written or solved correctly. This paper focuses on the MS handwritten recognition affected by noise during image acquisition. The architecture of the proposed algorithm is to create a deep neural network (DNN), to remove noise from a capture image using mathematical morphology operators (MMO) filters, and to check if the filtered image can be classified by our proposed DNN. Finally, the results of MS handwritten recognition after noise removal are shown and discussed.