Research On Online Overlaid Handwritten And Left-right Free Handwritten Word Segmentation

In this paper, we propose a novel string input mode based on the traditional handwritingway from left to right, which is named Left-Right Free Write. It not only support commonhandwriting way from left to right, but also supports handwriting from right to left andleft-right alternately handwriting. Others have proposed a string input mode: OverlaidHandwriting. It supports the phrase or sentence input function in mobile devices of a smallscreen size. It does this by writing a character to play down the track makes the whole processof writing coherent. Without affecting the user experience, it can accelerate the character inputspeed.Both Overlaid Handwriting and Left-Right Free Write are string entry mode. In the stringentry mode, it is very difficult to split the string. This paper presents a method of splitting ahandwriting string based on decision tree model and RBF neural network model, whichaddressing the string segmentation problem in Overlaid Handwriting and Left-Right FreeWrite.The main contribution of this paper is as follows:1. This paper presents three new evaluation factors about string segmentation, which isBCRV LTVCCR, N-PCR.2. The Overlaid Handwriting samples and Left-Right Free Write sample are automaticallygenerated from several single character databases through mathematical methods.3. This paper takes the framework of string segmentation and recognition. Whenever a newstroke is input, it is judged by a stroke classifier whether it is a new character componentor not. The stroke classifier is obtained by machine learning based on decision tree modeland RBF neural network model. Candidate character or several adjacent candidatecharacters is recognized by a character classifier. When the pen lift time exceeds athreshold, the recognition system assumes that the user finished writing. Experiments ona database of online Chinese handwriting demonstrate the effectiveness of the proposedapproach. The proposed segmentation approach achieved the recall rate of98.48%andthe effective rate of92.42%in Overlaid Handwriting at real-time mode. The proposedsegmentation method achieved the recall rate of99.08%and the effective rate of74.71% in Left-Right Free Write at real-time mode.