Bachelor Thesis - Computer Graphics Group
Bachelor Thesis - Computer Graphics Group
Bachelor Thesis - Computer Graphics Group
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Title: Mouse Gestures<br />
Author: Milan Burda<br />
Department: Department of Software and <strong>Computer</strong> Science Education<br />
Supervisor: RNDr. Josef Pelikán<br />
Supervisor’s e-mail address: Josef.Pelikan@mff.cuni.cz<br />
Abstract: In the presented work, we design and implement a mouse gesture<br />
recognition application. The program integrates transparently with the operating<br />
system, thus allowing existing unmodified Windows applications to be<br />
controlled by gestures. In an editor provided, the user is able to define a custom<br />
set of gesture patterns that the program automatically learns to recognize.<br />
The recognition algorithm is based on a preprocessing phase and two different<br />
gesture classifiers: back-propagating artificial neural network and k-nearest<br />
neighbors. The user is allowed to configure both general and application specific<br />
gesture mappings. These specify the commands to be triggered by the<br />
individual gestures. Several new features and improvements have been proposed<br />
for further development.<br />
Keywords: mouse gestures, gesture recognition, neural network, k-nearest<br />
neighbors, C++<br />
Název práce: Mouse Gestures<br />
Autor: Milan Burda<br />
Katedra (ústav): Kabinet software a v´yuky informatiky<br />
Vedoucí bakaláˇrské práce: RNDr. Josef Pelikán<br />
e-mail vedoucího: Josef.Pelikan@mff.cuni.cz<br />
Abstrakt: V predloˇzenej práci navrhneme a implementujeme aplikáciu na<br />
rozpoznávanie tzv. mouse gestures. Transparentná integrácia do operačného<br />
systému umoˇzňuje gestami ovládaˇt existujúce aplikácie pre Windows bez ak´ychko<br />
ˇ lvek úprav. Uˇzívate ˇ l si v poskytnutom editore definuje vlastnú mnoˇzinu<br />
vzorov, ktoré sa program naučí automaticky rozpoznávaˇt. Rozpoznávací algoritmus<br />
je zaloˇzen´y na fáze predprípravy a dvoch rôznych klasifikátoroch:<br />
neurónová sieˇt a k-najbliˇzˇsích susedov. Uˇzívate ˇ l si môˇze nadefinovaˇt vˇseobecné,<br />
ako aj ˇspecifické mapovanie gest pre rôzne aplikácie. Tieto mapovania definujú<br />
príkazy aktivované dan´ym gestom. Pre budúci v´yvoj bolo navrhnut´ych nieko ˇ lko<br />
nov´ych vlastností a vylepˇsení.<br />
Klíčová slova: mouse gestures, rozpoznávanie gest, neurónová sieˇt, k-najbliˇzˇsích<br />
susedov, C++<br />
5