• 1
  • 2
  • 5
  • 6
  • 3
  • 4

“School of Cognitive Sciences”

Back to Papers Home
Back to Papers of School of Cognitive Sciences

Paper   IPM / Cognitive Sciences / 9572
School of Cognitive Sciences
  Title:   Eye Movement Prediction in Visual Interactive Environments Considering Physical Actions
  Author(s):  Ali Borji
  Status:   In Proceedings
  Proceeding: AVA Annual Meeting, Manchester University, UK
  Year:  2008
  Supported by:  IPM
Human and machine vision systems are limited in terms of processing huge amounts of incoming visual information. Visual attention proposes a solution to such limitation by selecting those visual areas worth further higher cognitive processing. Eye movements convey much information about overt visual attention and are believed to be task dependent (Yarbus, 1967). Recent studies have been focused on prediction of eye movements in visual interactive environments rather than static synthetic search arrays (Peters et al, 2007, ACM trans,2, 1-21 ). In this study, we aim to investigate the role of previous eye movements and physical actions on future eye positions. Each frame and its associated physical action and eye position were recorded while subjects played video games (Nintendo, Need for speed). An augmented vector of fourier components representing each frame and its associated action were passed as input to a standard classifier (Multi-layer neural networks, Support-vector machines). The classifiers were trained to find associations of these vectors and their corresponding eye positions. Learned classifiers over 23 frames of each second of video a game was later tested over the remaining test frames. Our results show high classification rates for prediction of eye movements in visual interactive environments. They also suggest evidence toward usefulness of previous physical actions on prediction of future eye movements.

Download TeX format
back to top
Clients Logo
Clients Logo
Clients Logo
Clients Logo
Clients Logo
Clients Logo
Clients Logo
Clients Logo
scroll left or right