Open AccessAssamese Handwritten Character Recognition using Supervised Fuzzy Logic
Author(s) -
Kalyanbrat Medhi,
Sanjeeb Kalita
Publication year - 2020
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.d9989.018520
Subject(s) - pattern recognition (psychology) , artificial intelligence , preprocessor , fuzzy logic , numeral system , computer science , character (mathematics) , speech recognition , mathematics , geometry
This paper presents a state of the art supervised fuzzy pattern recognition system for recognition of Assamese handwritten characters. The fuzzy classifier is well suited for applications with ambiguities and handwritten character recognition is such a task. The dataset used in this experiment is taken from ISI Kolkata. After preprocessing images are normalized into uniform size 42x32 and then two features namely distance vector and density vector have been extracted. The experiment has two stages, training and testing. In first stage we extract distance vector and density features from uniform zones of the binary images for training classes and estimate the mean and variance for each class. In second stage we use this mean and variance to calculate the membership values for each unknown character of the testing set of data. An exponential fuzzy membership function is used for this purpose. Finally we recognize an unknown test character as that class for which it gives highest membership value. Finally result is stored in editable document. The highest recognition accuracy achieved in the experiment is 88.29%, 86.55% and 82.74% for numerals, vowels and consonants respectively.