Janez Bindas (2009) Automatic handwriting recognition on tablet personal computer. EngD thesis.
The thesis addresses the question of automatic handwritten recognition, applied to tablet personal computers. Handwritten recognition, as a special category of machine learning applications, is usually treated in the frame of pattern recognition research field. Automatic handwriting recognition, as a part of character recognition problems, has been a subject of research for more than 40 years. The performance of humans in text recognition has been a major driving force behind many research activities. Namely, this field is quite important from the scientific, commercial, engineering, application oriented and socially oriented point of view. For the purpose of automatic handwriting recognition, many methods of on-line and off-line recognition methods and techniques has been developed. Each method has indicative advantages and weaknesses, where an implementation of individual method depends on the nature of treated text and the problem, related to the recognition needs. Like other recognition problems, automatic handwritten recognition, applied to tablet personal computers, can be considered as an interesting and very intellectually challenging problem. This can be also evident from the rising number of publications, related to the intensive research activities in the field of character recognition, number recognition, greek and other symbols recognition, and even recognition of mathematical expressions and equations. The main objective of this thesis is to represent a new algorthm for an on-line tablet PC handwritten recognition, which is based on the two-stage recognizing concept. In the first stage, the algorithm deals with the training process, where recognizer is supplied with the knowledge about handwritting. More precisely, in the training stage, a character based reference model is constructed, where a set of typical attributes for each individual character is generated simultaneously with the handwritting process. In the second stage, an induced trained algorithm can be used for the purpose of practical implementation of handwritting recognition on the tablet PC. The main concepts of all crucial mechanisms, which are contained in the algorithm, are introduced in this work. Moreover, all important sub-algorithms of the main algorithm are carefully explained in the form of block shemes.
Actions (login required)