An Efficient Way of Handwritten English Word Recognition

Abstract

Handwriting recognition has been one of the most fascinating and challenging research areas in the field of image processing and pattern recognition in the recent years, which is motivated by the fact that for severely degraded documents the segmentation based approach will produce very poor recognition rate. The quality of the original documents does not allow one to recognize them with high accuracy. Hence, the aim of this research is to produce system that will allow successful recognition of handwritten words, which is proven to be feasible even in noisy environments. This paper presents a method that performs pre-processing steps on hand written images such as skew and slant correction, baseline estimation, horizontal and vertical scaling. It uses structural features for feature extraction. Further, Euclidean distance method is applied for classification that produces single matching word having minimum difference value. This paper presents a sample of data set which encompasses the names of 30 districts present in the Karnataka state of India. This method is useful for the postal address, script recognition and systems which require handwriting data entry.