Paper Title
Offline Handwritten Gurumukhi Place Names Recognition using Curve Fitting Based Features

Abstract
In the field of document analysis and recognition, postal automation plays a significant role which interprets the addresses by recognizing handwritten words. This paper is an effort in this direction in which an approach to recognize offline handwritten Gurumukhi place names (words) is proposed using a holistic approach which recognize the whole word without explicit segmentation. Two features, namely, parabola curve fitting and power curve fitting based features are extracted from the word images. For recognizing words based on extracting features, three classifiers, namely, decision tree, Multilayer Perceptron (MLP) and random forest are employed. By evaluating the proposed approach on a dataset comprising 10,000 Gurumukhi place names, Random forest classifier achieved a recognition rate of 86.26% based on the power curve fitting based features with 5-fold cross validation technique. Based on recognition rates, it has been concluded that random forest classifier outperforms other considered classifiers. Keywords - Offline Handwriting; Word Recognition; Gurumukhi Script; Holistic Approach