A scheme for applying emerging computer vision technologies to interactive e-learning

Xu, Richard Yi Da

A scheme for applying emerging computer vision technologies to interactive e-learning

Xu, Richard Yi Da

Permalink

Publication Type:: Thesis
Issue Date:: 2005

Closed Access

	Filename	Description	Size
	01Front.pdf	contents and abstract	732.76 kB	Adobe PDF	View/Open
	02Whole.pdf	thesis	138.65 MB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Xu, Richard Yi Da
dc.date.accessioned	2017-01-23T02:57:25Z
dc.date.available	2017-01-23T02:57:25Z
dc.date.issued	2005
dc.identifier.uri	http://hdl.handle.net/10453/75645
dc.description	University of Technology, Sydney. Faculty of Information Technology.	en_AU
dc.description	NO FULL TEXT AVAILABLE. Access is restricted indefinitely. The hardcopy may be available for consultation at the UTS Library.
dc.description.abstract	NO FULL TEXT AVAILABLE. Access is restricted indefinitely. ----- In traditional e-learning applications, instructional video capturing is used merely for streaming purposes. The recent advancements in video signal processing, computer vision and pattern recognition (CVPR) technologies have led to instructional video also becoming the input for many automated and intelligent e-learning systems. At the same time, the increases in consumer-grade PC processing speed and camera resolutions have made CVPR technologies more accessible to ordinary computer users and have the potential to apply to their everyday learning. Noticing this trend, in our work, we have focused our attention on e-learning systems that are synchronous, peer-to-peer based and instructor uses traditional teaching equipment, where the properties include accurate bidirectional interactions, low-cost hardware, system portability and vision software versatilities are emphasized Our research also indicates that existing vision-based e-learning systems have not focused on these areas and fulfilled the corresponding requirements adequately. Our work, therefore, is an effort to address the challenges brought by these properties and to apply our computer vision research findings to help resolve the challenges. In this thesis, the work is presented through a scheme of attained results. Categorically, these results are presented from the perspectives of interactive multimedia systems, e-learning computer vision algorithms and processing implementations. From a multimedia system perspective, we have developed a few interactive subsystems where our work has represented both novelty and successful attention to our challenges. Our work includes a semi-passive control subsystem for a low-cost and low-precision PTZ camera able to be operated according to both instructor-led laser pointer guidance and by a student in a remote location. We also present a novel student-to-instructor communication method using a low-cost laser pointer robot that can automatically pan and tilt to a specified region. We also present a real-time teaching object recognition subsystem, where both hand-held and wall-based teaching objects can be detected during synchronous e-learning. By using unique instructor's facial pose estimation and laser pointer guidance techniques, this system has achieved robustness, efficiencies and unambiguous detection. From a computer vision algorithms perspective, we will present results in researching and customizing various emerging CVPR algorithms to suit e-learning applications. These results include robust object recognition using local invariant features under reference scales; robust laser pointer detection by bimodal Gaussian Mixture Model training with integral images features; Application of mean-shift tracking to e-learning and safeguarded with efficient global target regaining. In these algorithms, the efficiency for single PC execution is also emphasized. From a processing implementation perspective, we will present results using a latency-insensitive CVPR scheduling method by our unique execution parameter generation functions for the expensive vision algorithms used. We also present a method for peer-to-peer CVPR processing load-sharing on student's PC. SIMD optimization results and studies into automatic SIMD compilation on vision algorithms are also discussed. Finally, the overall results obtained are presented though our prototype IVDA instructional video streaming system, encapsulating most of our work. Apart from presenting a state-of-art vision-based e-learning system, we will also illustrate how the results obtained have addressed and resolved the challenges in our focused e-learning areas.	en_AU
dc.format	Thesis (PhD)
dc.language.iso	en_AU	en_AU
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	info:eu-repo/semantics/closedAccess
dc.title	A scheme for applying emerging computer vision technologies to interactive e-learning	en_AU
dc.type	Thesis	en_AU
utslib.copyright.status	closed_access

Abstract:

NO FULL TEXT AVAILABLE. Access is restricted indefinitely. ----- In traditional e-learning applications, instructional video capturing is used merely for streaming purposes. The recent advancements in video signal processing, computer vision and pattern recognition (CVPR) technologies have led to instructional video also becoming the input for many automated and intelligent e-learning systems. At the same time, the increases in consumer-grade PC processing speed and camera resolutions have made CVPR technologies more accessible to ordinary computer users and have the potential to apply to their everyday learning. Noticing this trend, in our work, we have focused our attention on e-learning systems that are synchronous, peer-to-peer based and instructor uses traditional teaching equipment, where the properties include accurate bidirectional interactions, low-cost hardware, system portability and vision software versatilities are emphasized Our research also indicates that existing vision-based e-learning systems have not focused on these areas and fulfilled the corresponding requirements adequately. Our work, therefore, is an effort to address the challenges brought by these properties and to apply our computer vision research findings to help resolve the challenges. In this thesis, the work is presented through a scheme of attained results. Categorically, these results are presented from the perspectives of interactive multimedia systems, e-learning computer vision algorithms and processing implementations. From a multimedia system perspective, we have developed a few interactive subsystems where our work has represented both novelty and successful attention to our challenges. Our work includes a semi-passive control subsystem for a low-cost and low-precision PTZ camera able to be operated according to both instructor-led laser pointer guidance and by a student in a remote location. We also present a novel student-to-instructor communication method using a low-cost laser pointer robot that can automatically pan and tilt to a specified region. We also present a real-time teaching object recognition subsystem, where both hand-held and wall-based teaching objects can be detected during synchronous e-learning. By using unique instructor's facial pose estimation and laser pointer guidance techniques, this system has achieved robustness, efficiencies and unambiguous detection. From a computer vision algorithms perspective, we will present results in researching and customizing various emerging CVPR algorithms to suit e-learning applications. These results include robust object recognition using local invariant features under reference scales; robust laser pointer detection by bimodal Gaussian Mixture Model training with integral images features; Application of mean-shift tracking to e-learning and safeguarded with efficient global target regaining. In these algorithms, the efficiency for single PC execution is also emphasized. From a processing implementation perspective, we will present results using a latency-insensitive CVPR scheduling method by our unique execution parameter generation functions for the expensive vision algorithms used. We also present a method for peer-to-peer CVPR processing load-sharing on student's PC. SIMD optimization results and studies into automatic SIMD compilation on vision algorithms are also discussed. Finally, the overall results obtained are presented though our prototype IVDA instructional video streaming system, encapsulating most of our work. Apart from presenting a state-of-art vision-based e-learning system, we will also illustrate how the results obtained have addressed and resolved the challenges in our focused e-learning areas.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/75645