Online handwriting analysis is achallenging research area for the last few decades. To recognize with perfectprediction some pre-processing steps are essential. In this paper one of theimportant pre-processing steps dehooking is presented.
Here we consider Bengalionline handwritten Characters and words as sample for removing hooks. Hooks arebasically common artifacts during fast writing by people. Hooks are very commonissue present at the beginning in very rare case and the end ofcharacter stroke in maximum case and are generated by the pen-down and pen upmovements respectively. Dehooking is the process of eliminating such unwanted strokes that appeardue to inaccuracies in pen down position.
Dehooking algorithms are applied toremove hooks. Here, strokes are detected by comparing the number of points witha threshold value. If the value is greater than the threshold value, the markis retained or it is removed otherwise. In this new and innovative approach we focus on thedehooking at the end of character stroke and consider last 20 percent of eachstroke for checking, according to distance from the co-ordinate of the firstpixel. In last 20 percent of a stroke, we calculated angle among threeconsecutive pixels. If in a particular point, angle among three consecutive pixelsis falling suddenly then immediately we pointed out that point.
After pointingout the angle falling place we checked the entire remaining pixel after thatpoint, whether all the remaining points are getting fade slowly or not. If itis found that all the remaining points aregetting fade slowly then it can beassumed that it a hook. After detecting the hook for a particular stroke we removedall the remaining pixels from the falling angle place so that hook can beremoved and the handwritten character will be hook less.I tested 1200 Englishand 1600 Bengali online handwritten characters and we got 97.
02 percent ofaccuracy.Online Handwriting recognition is a procedureof a computer to detect and understand intelligible handwritten characters,words, sentence or paragraph input from a touch sensitive or pen sensitiveinput sources such as Pen tablets, PDA, touch-screens and other devices. Themovements of the pen tip may be sensed “on line”, but it iscomparatively difficult task to recognize with great accuracy because in caseof online handwriting only co-ordinate values are known to us. Handwritingrecognition principally entails optical character recognition. However, acomplete handwriting recognition system also handles pre-processing steps,formatting, performs correct segmentation into characters, normalization andfinds the most plausible Characters and words. On-line handwriting recognitioninvolves the automatic conversion of text as it is written on a specialdigitizer or PDA, where a sensor picks up the pen-tip movements as well aspen-up/pen-down switching and covert into vectors or matrix form.
This kind ofdata is known as digital ink and can be regarded as a digital representation ofhandwriting. The obtained signal is converted into letter codes which areusable within computer and text-processing applications.