An Efficient Off-line Handwritten English Alphabet Character Recognition based on Hidden Markov Model and Discrete Wavelet Transform

Abstract

An algorithm for the recognition of handwritten English alphabet characters based on Hidden Markov models (HMMs) and Discrete Wavelet Transform (DWT) is proposed. In order to additional reduction in computational complexity and memory consumption, the handwritten English alphabet character images are resized to 9x16 bmp format using Discrete Wavelet Transform and Universe of Discourse techniques. The features incorporated in the HMMs are based on the geometrical properties of the characters. After extracting zones/blocks from each of the preprocessed handwritten English alphabet character, each zone/block is transformed into a 09-dimensional feature vector which is further reduced to 05 using SVD technique to reduce the computational complexity. The geometrical features represent information about the number of horizontal lines, vertical lines, left diagonal line, right diagonal lines and normalized number of horizontal lines incident to the considered zone/block. Each letter of the alphabet is represented by an HMM. Training of the HMMs is done by means of the Baum-Welch algorithm, while the Viterbi algorithm is used for recognition. An average comparable recognition rate of over 96.15% on the character level has been achieved at reduced computational cost of 0.8%.