An Enhanced Algorithm with Combined Feature Extraction and Neural Network Approach for Recognition of Handwritten Mathematical Expressions

Identification of math symbols and expressions is an inspired task due to its 2D layout, complex semantics and spatial structure. Extracting math expressions from document is more tangled. Actually many approaches have been proposed for better recognition of math expressions but does not resolved ambiguities in some cases of complex handwritten mathematical expressions. In the proposed recognition system, fusion of machine learning and combined feature extraction (boundary box, counting elements, height to width ration, and area of elements) is evaluated by feed forward back propagation neural network. The experimentation and analysis has been carried out for various kinds of handwritten mathematical expressions. The difficulties in recognition of complex handwritten mathematical expressions have been solved by training the neural network with scaled conjugate gradient in proposed methodology. The system verifies its strength and potential. The proposed method enhanced the speed of recognition with expecting results and confusion matrix shows the accuracy of 98.6%.