This thesis explores perceptual breakthrough in rapid serial visual presentation (RSVP), for deception detection applications. In RSVP, visual stimuli are presented in rapid succession, pushing the perceptual processing system to the limit, allowing only a limited number of stimuli to be processed and en- coded. In this thesis we investigate what type of stimuli capture attention in RSVP, taking advantage of both physiological and behavioural measurements. The main focus of the studies presented here follows up on work that shows that perceptual breakthrough in RSVP can be used as a marker of concealed knowledge in deception detection tests (Fringe P300). The thesis is divided into two research contribution parts. Firstly, we develop methods for analysing Event Related Potential (ERP) data, in order to facilitate assessment of perceptual breakthrough in experiments presented later in this thesis. We focus on reducing false positives while at the same time successfully measuring the underlying effects. We present and evaluate methods for measuring latencies and selecting Regions of Interest (ROIs) through simulations and experimental data. Secondly, we explore perceptual breakthrough in RSVP with applications in deception detection. For that purpose, we conducted two studies. The first study explores incidentally acquired information by recording the P300 ERP component from participants after acting out a mock crime scenario. The main hypothesis was that concealed information is salient to a guilty person, and thus associated stimuli will be involuntary perceived. The second study explores the type of stimuli that capture attention in RSVP, by addressing issues related to encoding and emotional arousal, and whether attention can be directed through contextual priming independent of the main task. These studies increase our understanding of how stimuli are processed in RSVP and can provide useful suggestions for designing more successful ERP and RSVP based, deception detection applications, both in terms of stimulus presentation and data analysis.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:713020 |
| Date | January 2016 |
| Creators | Zoumpoulaki, Alexia |
| Contributors | Bowman, Howard |
| Publisher | University of Kent |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://kar.kent.ac.uk/61386/ |
Page generated in 0.0033 seconds