Hindi Online Handwritten Word Recognition

Anand Sharma and A G Ramakrishnan

Online Handwritten Word (OHW)

Online handwritten word is generated as X-Y coordinates of pen like device applied on touch sensitive screen. X-Y coordinates of pen tip are recorded as time dependent signal. Pen-down state occurs when the pen touches the screen and pen-up state occurs when pen is lifted from the screen. Sequence of X-Y coordinates between pen-down and pen-up states is called a stroke. Group of strokes forms a character and group of such characters forms a word.

Devanagari Character Set

Characters from the Devanagari Script

Hindi language is written using Devanagari script. Some Devanagari characters are as shown below.

Online Handwritten Word Recognition

An individual will write a character differently at different times. So a word consisting of many characters written by different people will have a lot of variability in it. Objective of online handwritten word recognition algorithm is to recognize online handwritten words written by different people by correctly recognizing the characters in the word.

Block Diagram of Word Recognition System

The word recognition algorithm being studied is unconstrained and writer independent and recognizes a word by segmenting the word and classifying the characters in the word. Character classifier is built by training the classifier on set of online handwritten samples of characters. Segmentation of word is done by grouping strokes based character class label confidence. This combination of segmentation and recognition process produces a sequence of classified characters. The word is considered recognized in terms of classified characters constituting the word.