Bayesian recognition of human identities from continuous visual features for safe and secure access in healthcare environments

When using today's admission control systems some kind of interaction is generally required. That is because the system needs something to check a person's identity before it allows access. Often the system will ask for a pin or a password. For many people these interactions are a problem as they might be unable interact with the system in the required way due to impairments. This can be caused by a disease such as Parkinson's or Alzheimer's or they might be unable to use their hands for other reasons. In hospitals, retirement homes or assisted living communities many people would benefit from a contactless admission control system. We propose a system that recognizes human identities from continuous geometrical features that were estimated using the Kinect 2. Those include the distance between the eyes and the estimated height of a person. A Bayesian Classifier for continuous features is used to predict people's identities. The preliminary recognition results suggest that the proposed geometrical features can contribute to reliable admission control. This work was supported by SysTeam GmbH Dortmund, Germany.