Offline handwritten gujarati word recognition

Abstract

Gujarati language is an Indo-Aryan language that has a complex structure wherein extracting each character becomes hectic because of the presence of diacritics. Implementation of word recognition technique on the Gujarati database makes the work easy as it does not require extraction of symbols and glyphs from the word image. In this paper, we describe a handwritten Gujarati word recognition technique using Histogram of Oriented Gradients (HoG) features and state of the art classifier like Support Vector Machine (SVM) and k-Nearest Neighbor (kNN). The experiments were performed on a moderate sized database of handwritten Gujarati city names. The work produced has direct application in handwritten postal address processing.