Optical Character Recognition for Handwritten Mathematical Expressions in Educational Humanoid Robots

Abstract

Humanoid robots are seen as perfect education partners for students these days. Many researchers are working on educational humanoid robots to make them more effective in educating students. One of many approaches to realize it is to make the robots educate children in two-way communication. In this case, robots can recognize students' progress and provide suitable learning materials according to their progress. In a simple scenario for basic mathematics learning, the robot educates a student on how to solve mathematical expressions. As the student understand the concepts, the robot gives some simple exercises to further improve student's ability. While completing the exercise, the partner robot supervises each mathematical expression solved by the student, acknowledges mistakes and provides correct solutions. Thus, the robot must be equipped with a system which can recognize mathematical expressions and give the solutions. Such system is an inherent foundation to build much smarter robots in delivering education to students. This paper covers detail design and implementation of the system which recognizes handwritten mathematical equations and provides the corresponding solutions. Testing results show that the system performs quite well with 97.4% accuracy with slight errors for characters with similar shapes.