A Survey on Recognition and Analysis of Handwritten Document

Handwriting differs from person to person. Some may be legible while some others are difficult to read or understand. Hence this project aims at recognizing the handwritten text and understanding what it is with the help of a neural network and fuzzy logic. It involves segmentation, feature extraction and classification.Here the method used is Canny Edge Detection Algorithm and the Histogram Of Gradients for the feature extraction. The neural network is trained on to a 50 set samples for each of the 26 alphabets and 10 numbers for recognition. The fuzzification can be applied along with this inorder to get more accurate results by giving the questionnaires, ie, by giving the conditions to check if it satisfies a particular character which is to be determined. This would thus yield 80 percentage accuracy and reliability in recognition of the handwritten text. Keywords— Neural network, Histogram of Gradients, Canny algorithm, fuzzy logic