Gaze-Aware Cognitive Assistant for Multiscreen Surveillance

Abstract

Surveillance operators must scan multiple camera feeds to ensure timely detection of incidents; however, variability in scanning behavior can lead to untimely/failed detection of critical information in feeds that were neglected for a long period. Us-ing an eye tracker to monitor screen fixations we can calculate (in real-time) the time elapsed since the last scan of each particular feed, allowing the setting-up of targeted countermeasures contingent on operator oculomotor behavior. One ave-nue is to provide operators with timely alerts to modulate the scan pattern to avoid attentional tunneling and inattentional blindness. We test such an adaptive solution within a major event surveillance simulation and preliminary results show that operator scan behavior can be modulated, although further investigation is re-quired to determine warning frequency and modality to optimize the balance be-tween saliency and workload increase. Future work will focus on adding a real-time vigilance detection and countermeasure capability.