A Combined Approach of Feature Selection and Machine Learning Technique for Handwritten Character Recognition

Abstract

Handwritten Character Recognition has proven its worth in most challenging research areas in the various application based domains like bank cheques, signature verification, aids for blind persons, etc. The role of alphabetical characters is quite focused to recognize the scanned text that can be further deployed for sentimental analysis. In this research paper, we have deployed a promising approach to recognize handwritten English alphabets. We have mainly aimed in the digits recognition. The features of Handwritten Numerals are extracted using geometrical and gradient-based measures. The redundancy of feature set is reduced using feature selection method, and machine learning approach is deployed to generalize the decision boundary. The experimental outcome proves its promising worth comparable to existing arts.

Keywords: geometrical features, gradient feature, selection, cross-validation decision, boundary