Abstract

Recognition of handwritten digits, characters, mathematical symbols and equations is an intricate task due to its 2\ndimensional layout, variation in writing style, different font, shapes, complex semantics, and spatial structure. Extracting\nmathematical equations from the scan document is more complex. The proposed recognition system completes the task by\nusing feed forward back propagation neural network and hybrid feature extraction technique. The experiment has been\ncarried out for different types of handwritten mathematical equations. The system verifies its accuracy. By using neural\nnetwork with scaled conjugate gradient training, the accuracy increases with enhancing the speed of recognition.