Affective user threat profiling using computer vision-based heart rate estimation in profile-based surveillance environments

In current public spaces, it is difficult to maintain security and determine if someone is a potential threat. In most cases, reactive approaches are applied after an event occurs to mitigate the threat, rather than being proactive and attempt to prevent the event from occurring in the first place. Modern CCTV camera surveillance can efficiently track users and match against a watch list for potential threats, but it is up to the surveillance operator to analyse user behaviour that may classify a user as a threat. However, this approach is subject to human error and requires human resources to facilitate it. This article proposes a technology to automate this process by using the existing camera infrastructure and using affective computing methods for user threat profiling. The preliminary results show that certain environments that allow for a profile view, yield good classification results. Although there are certain structural and environmental constraints, the technology is viable and warrants further investigation.