*Reliability and testing in vision-based interaction.
Item
-
Title
-
*Reliability and testing in vision-based interaction.
-
Identifier
-
AAI3213145
-
identifier
-
3213145
-
Creator
-
Mbogho, Audrey J. W.
-
Contributor
-
Adviser: Lori L. Scarlatos
-
Date
-
2006
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Computer Science
-
Abstract
-
While reliability is of crucial importance in software engineering in general, it is especially so in human computer interaction (HCI), where the success or failure of a product hinges upon the level of confidence users have in it. This confidence can be increased by a product's demonstrated reliability. On the other hand, it can be quickly corroded when the product behaves in unexpected ways. Software testing is the most effective approach for revealing weaknesses in software so that they can be eliminated and the software made more reliable. But software is rapidly changing, with current trends in human computer interaction leaning towards perceptual interfaces, which accept real-time, complex, and uncertain input data, such as those from cameras, microphones, and tablets. In particular, on-going research activities in computer vision and HCI seem to indicate that camera-based inputs are poised to play an important role in human computer interaction. Furthermore, advances in camera technology and the appearance of inexpensive, consumer-grade products that integrate cameras into computing devices are strong indicators that vision-based interaction is imminent. There is a question as to whether established practices in software testing, where the test data are predictable and have known outputs, are ready for these trends. How can a vision-based system, given the uncertainty of its input data and the variability of its outputs, be tested so that it produces a result that demonstrates reliability and engenders user confidence?;This thesis highlights the challenges computer vision brings to the testing of software based on it. By looking at it as a parameter optimization problem, this work examines how approaches in traditional software testing can be augmented to adequately address uncertainty in building reliable vision-based interaction. A solution strategy which incorporates genetic generate-and-test techniques is proposed. The viability of this approach is demonstrated with experiments in which objects that are part of a tangible user interface for educational applications are augmented with visual tags. These visual tags are captured by camera and identified through image analysis. This research shows that findings in this domain can be applied in other domains.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
