Back to search

National Technical University of Athens (NTUA)

Αναγνώριση χειρόγραφων μαθηματικών εκφράσεων εντός γραμμής (online)

Abstract

dc:description

The thesis focuses on the recognition of online handwritten mathematical expressions. The main contribution is the development of a system for the recognition of online handwritten mathematical expressions. This system operates in three stages: a) grouping strokes to symbol hypotheses and recognizing these candidate symbols, b) recognition of the spatial relation between two corresponding mathematical symbols or subexpressions, and c) representation of the mathematical expression using the MathML format (Appendix B).In the first stage, we use the Freeman-8 chain code followed by the local percentage of the length between two successive points to represent the symbol. The Freeman-8 chain code is used to represent a boundary by a connected sequence of straight-line segments of specified length and direction. The direction of each segment is coded by using an integer number in the range of 0 to 8. For the comparison between the reference symbol and the test symbol, the template matching technique is used. In the second stage, the spatial relations between the symbols that comprise the mathematical expression are recognized. The geometrical features of the symbols comprising the mathematical expression (ME) (e.g. centroid, bounding box) are used in order to classify the relation between symbols (e.g. superscript, subscript) and finally put the symbols in the right subexpression of the ME.In the current thesis two different techniques are developed. In the first technique the geometrical classifier for the recognition of the spatial relation between symbols comprising the ME is used. After the spatial relation is recognized, the symbols are positioned in the corresponding levels of the ME. The symbols that belong to different levels are connected with symbols of other levels of the ME based on the spatial relation between them. Finally, when all the symbols are positioned in the corresponding levels and the spatial relations between the symbols are defined, a MathML file is produced for the representation of the ME. The mathematical expression that can be recognized through this technique is till one level up (superscript) and one level down (subscript). In the second technique an SVM based classifier is developed for the recognition of the spatial relation between symbols of the ME. In the third stage a stochastic context free grammar (SCFG) is used for the description of the syntax of the MEs. When the grammar is defined we can then search for the most likely ME using also the spatial relations between the symbols and/or subexpressions. To analyse effectively the ME, we first need to define a parsing algorithm for the ME. In the current thesis we use a variation of the Cocke–Younger–Kasami (CYK) algorithm in order to tackle the problem of the analysis of the MEs.Also a program for the acquisition of handwritten mathematical expressions is developed for the creation of a handwritten mathematical symbols dataset. With this system a database 186 symbols selected by 50 different writers was created. Each writer wrote each symbols 5 times. Moreover each writer wrote 54 mathematical equations that contained at least once each symbol of the dataset.We use two different datasets of mathematical symbols, the Laviola dataset and the CROHME dataset. The accuracy rate for the symbol recognition algorithm is 92% for the LaViola dataset and 77.25% for the CROHME2014 dataset. For the evaluation of the algorithm for the recognition of the spatial relation between objects, we used the MathBrush dataset. The reported error rate in this case is 2.87%. Finally the CROHME 2012 and 2013 datasets were used for the evaluation of the parsing algorithm of the MEs and gave an accuracy result of 78.70%, 65.78%, 56.37% and 50.22% for the four different parts of the dataset.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2015

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Simistira, Foteini
  • Σιμιστήρα, Φωτεινή

Subjects

dc:subject × 17

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/39800

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Simistira, Foteini; Σιμιστήρα, Φωτεινή. Αναγνώριση χειρόγραφων μαθηματικών εκφράσεων εντός γραμμής (online). National Technical University of Athens (NTUA), 2015. http://hdl.handle.net/10442/hedi/39800