Handwritten Recognition using Slope and Curvature Functions

Abstract

Letter recognition and handwritten processing is one of the major and open problems in Artificial Intelligent (AI) domain. This study introduces a method based on statistical and geometrical techniques to recognize handwritten digits and letters. These techniques use the fuzzy logic to create the vector curves. Inputs are online digits or letters and outputs are two arrays of slope and curvature values. The slope and curvature values of training data are stored in a database and used in comparison phase. The test results show that 96.98% of inputs are correctly recognized.