Handwritten Numeral Recognition Using Wavelet Transform and Neural Networks

Character recognition is one of the most popular practical applications of pattern recognition. The first step is to acquire a digital image. The next step is acquiring to improve it for the other processes. The next stage is segmentation, the pattern recognition problem can be represented into data acquisition, preprocessing, feature extraction, and classification. The wavelet transform is used in including image processing [1][2]. For a given image the wavelet transform produces a low frequency sub band image reflecting its basic shape and three sub band images that contain the high frequency components of the image at horizontal, vertical and diagonal directions. The artificial neural networks can be successfully used In this paper we will use wavelet coefficients as features in numeral recognition problem, these features are fed to MLP network which used as a classifier. We will train and test MLP networks by using different types of discrete wavelet transform which applied to a data base of 960 samples of handwritten numerals. The goal of this paper is to demonstrate the strengths of wavelet transform as a tool of feature extraction combined with a neural network as a classifier in the field of numeral recognition