SURVEY ON HANDWRITTEN DEVNAGARI CHARACTER RECOGNITION

Character Recognition by machines is an innovative way by which the dependence on manpower is reduced. Character recognition provides a reliable alternative of converting manual text into digitized format.Now-a-days, as technology becomes integral part of human life, many applications have enabled the incorporation of English OCR for real time inputs. The advantages that the English alphabet has is its simplicity offered by less number of lettersi.e. 26 and easier classification due to the concept of lowercase and uppercase. If we consider Devnagari script in this scenario, we will come across myriad hurdles because this script lacks the simplicity of English. The concept of fused letters, modifiers, shirorekha and spitting similarities in some letters make recognition difficult. Also, character recognition for handwritten text isfar more complex than that for machine printed characters. This is because of theversatility and different writing techniques adopted by people. The direction of strokes, pressure applied on writing equipments, quality of writing equipment and the mentality of the writer itself highly affects the written text. These problemswhen combined with the intricate details of Devnagari script, the complications in constructing a HCR of this script are increased.The proposed system focuses on these two issues byadopting Hough transform for detecting features from lines and curves. Further,for classification, SVM is used. These two methods when combinedprovide high accuracy which is up to 90%. Prior to these techniques, pre-processing ofcharacters is done to ensure accurate classification. This system is highly useful as it canbe used for automation of various services like postal, rail etc.