Online Handwritten Mathematical Expression Solver Using Artificial Neural Network

Abstract

Mathematics plays a predominant role in each of our lives. When it comes to solving a mathematical expression, we are highly dependent on the calculators that are available in almost every electronic gadget. Since all these gadgets are touchscreen-based nowadays, building a system that recognizes and solves online handwritten mathematical expressions is the potential area of this research. Recognition of online handwritten mathematical expressions is a complicated task. In this paper, an Artificial Neural Network model is built for the recognition of handwritten digits, operators, and symbols. Tkinter GUI interface is built for the users to type in their expressions and image processing is done by capturing an image from the canvas and converting it into a NumPy array and then applying the thresholding technique to convert it into a binary array. Connected component labeling is done to separate every number and symbol on the canvas. These numbers and symbols are then sent to the artificial neural network for predictions. The model gave a training accuracy of 98.97% and a test accuracy of 98.95%. Finally, the expression is evaluated, and the translated expression and output are shown on the Tkinter GUI interface.