We are investigating what drives human attention and eye-movements when observing moving images. Our goal is to develop a comprehensive theory of active visual perception and cognition. Questions which motivate this research include:
* How do we look at film, television, on-line video or real-life events?
* How does looking influence what we see, remember and feel?
Using state-of-the-art eyetracking facilities, the eye-tracking data we includes the X and Y coordinates for each eye per millisecond in combination with a range of eye-movement measures. For each person, this produces over 1000 lines of data (in eight columns) per second! To analyze this output, we have developed a tool called Computational Algorithms for Representation and Processing of Eye-movements (CARPE). CARPE can create dynamic heat-maps (like the image in the top left corner of this page) which visually represent a cluster analysis of where people are looking. These results are shown in our videos, which were produced by CARPE. The software also quantifies this behavior, allowing us to determine the various visual features and events that lead to a stronger consensus on where to look (a tighter cluster) among viewers.
The DIEM database
The DIEM database can be found here: DIEM Database. Over two hundred and fifty volunteers participated in the project. The corpus offers insights relevant to a range of academic and end-users, covering topics such as psychology, computer vision, advertising and film studies.
The DIEM Project was initially funded by the Leverhulme Trust and the Economic and Social Research Council (John M. Henderson, Principal Investigator). John Henderson is now at the University of California, Davis.