Implementation of Online Handwriting Recognition System for Ethiopic Character Set

Abera, Abebaw (2007) Implementation of Online Handwriting Recognition System for Ethiopic Character Set. Masters thesis, Addis Ababa University.

PDF (Implementation of Online Handwriting Recognition System for Ethiopic Character Set) Abera, Abebaw.pdf - Accepted Version Restricted to Repository staff only Download (505kB) | Request a copy

Abstract

Computers are now becoming involved in day to day activities of people. Besides the functionalities they provide, people are now seeking computers that are easier to use. Needs are growing to use computers everywhere and at any time, where people are located. This necessitates making computers pocket size to carry them while we are moving around. Handheld devices are pocket-sized devices that are now becoming widely used computers. Among the various types of handheld devices, PDAs are designed to have touch screen and stylus (pen-like device) that can facilitate the development of handwriting recognition system. While there has been other means of communication with computers, speech recognition and handwriting recognition are now becoming choices due to the fact that they are more natural. Much effort is required for these data input mechanisms so that users are expected to make only minimal corrections in using such systems. This project work performs the implementation of online handwriting recognition system for Ethiopic character set. Basically, the system is a writer-dependent system in which the user is expected to train the system. As the system is stroke number and order dependent, the user is obliged to provide his/her mostly used handwriting style during the training phase. In using the system after the training is completed, the user has to remember the order and number of strokes of the character in which the system is being trained. Missing to match the number of strokes of the input pattern with the training data will not result to a correct recognition by any means. A prototype is developed to show the practicality of the algorithm designed in [4]. In general, the system is developed to recognize the 34 basic characters and the remaining non-basic ones. For the non-basic characters a technique proposed in [4] is used and its practicality is shown by implementing for the second and third non-basic characters. Similarly the technique works for the remaining non-basic characters. Adding the non-basic characters to be recognized in this way does not have an effect on the recognition rate achieved for the basic characters. Prior to the prototype development, the algorithms designed in [4] are thoroughly analyzed and the design goals are set. To improve the efficiency of the system, the 34 basic characters are classified according to their number of strokes. Hence, those having the same number of strokes are grouped together. Additionally, the reference file organization is modified and the X and Y observation code sequences are made to be stored together in a single file. Due to the grouping of characters and the reference file organization change, originally designed algorithms are modified. Additionally, a new algorithm is designed for the data collection process. The superimposition algorithm, which is part of the recognition process, was designed to operate on the data points rather than the observation code sequences. This is computationally expensive and will make the system to store the data points persistently, besides the code sequences. Hence, the superimposition algorithm is not included in the prototype development. Java 2 Micro Edition (J2ME) programming language is used for the prototype development. To compile and preverify the source code Java Wireless Toolkit (JWTK) development environment is used. Since it was not possible for us to get a development environment that supports PDA emulators, debugging and testing errors and locating them was the most challenging part of this prototype development. Experiment is conducted on both the emulator and the PDA device, to test the system’s recognition rate. The experimental result shows that accuracy rates of 89.75% and 86.00% are achieved on the emulator and PDA device respectively.

Item Type:	Thesis (Masters)
Subjects:	P Language and Literature > P Philology. Linguistics Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software T Technology > T Technology (General)
Divisions:	Africana
Depositing User:	Selom Ghislain
Date Deposited:	18 Jun 2018 09:39
Last Modified:	18 Jun 2018 09:39
URI:	http://thesisbank.jhia.ac.ke/id/eprint/4279

Actions (login required)

View Item