Image sequence event detection VIA recurrence analysis

Recurrence analysis methods have been used in a wide array of fields for the purposes of obtaining some grasp of the characteristics of a chaotically dynamical system. At the heart of the recurrence plotting and quantification analysis algorithm, though, is a means of visualizing and measuring repeating sequences of time-dependent data. It is in this sense we present a novel means of extracting an event from an image sequence by analyzing selected features as multi-dimensional samples of the time-series. The analysis and development of recurrence plots (RPs) naturally lend themselves to sequence detection, and we are presenting an attempt to target eye-movement events as such time-series sequences. We can then apply relatively simple quantification measures through Recurrence Quantification Analysis (RQA), thereby immediately detecting an event and capturing some characteristic statistics. To highlight an application of this methodology, we are presenting an approach towards detecting fixational eye-movement events from a video recording of natural eye motions.