liu.seSearch for publications in DiVA
Change search
ReferencesLink to record
Permanent link

Direct link
Automatic Visual Behavior Analysis
Linköping University, Department of Electrical Engineering.
2002 (English)Independent thesis Basic level (professional degree)Student thesisAlternative title
Automatic Visual Behavior Analysis (Swedish)
Abstract [en]

This work explores the possibilities of robust, noise adaptive and automatic segmentation of driver eye movements into comparable quantities as defined in the ISO 15007 and SAE J2396 standards for in-vehicle visual demand measurements. Driver eye movements have many potential applications, from the detection of driver distraction, drowsiness and mental workload, to the optimization of in-vehicle HMIs. This work focuses on SeeingMachines head and eye-tracking system SleepyHead (or FaceLAB), but is applicable to data from other similar eye-tracking systems. A robust and noise adaptive hybrid algorithm, based on two different change detection protocols and facts about eye-physiology, has been developed. The algorithm has been validated against data, video transcribed according to the ISO/SAE standards. This approach was highly successful, revealing correlations in the region of 0.999 between analysis types i.e. video transcription and the analysis developed in this work. Also, a real-time segmentation algorithm, with a unique initialization fefature, has been developed and validated based on the same approach.

This work enables real-time in-vehicle systems, based on driver eye-movements, to be developed and tested in real driving conditions. Furthermore, it has augmented FaceLAB by providing a tool that can easily be used when analysis of eye movements are of interest e.g. HMI and ergonomics studies, analysis of warnings, driver workload estimation etc.

Place, publisher, year, edition, pages
Institutionen för systemteknik , 2002. , 45 p.
LiTH-ISY-Ex, 3259
Keyword [en]
Reglerteknik, visual behaviour, signal processing, saccade, fixation, smooth pursuit, eye tracking, head tracking, glance analysis, glance
Keyword [sv]
National Category
Control Engineering
URN: urn:nbn:se:liu:diva-1980OAI: diva2:19307
Available from: 2003-09-16 Created: 2003-09-16

Open Access in DiVA

fulltext(944 kB)1226 downloads
File information
File name FULLTEXT01.pdfFile size 944 kBChecksum SHA-1
Type fulltextMimetype application/pdf

By organisation
Department of Electrical Engineering
Control Engineering

Search outside of DiVA

GoogleGoogle Scholar
Total: 1226 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 612 hits
ReferencesLink to record
Permanent link

Direct link