Using Audio for Detecting Covered Users in Front of Public Displays

Abstract

Public displays have become more and more ubiquitous in recent years. Both for managing public display networks and for doing research in this field it is an important issue to automatically detect what is happening in front of these screens. This often is done with the help of depth cameras. However, measurement errors can occur due to occlusion or underexposure. In our research we looked into the idea to support the detection of concealed persons using audio information. A series of laboratory experiments showed that audio recordings can be used to detect concealed persons. This can be used in existing installations to minimize occlusion error in visual recordings.