Indoor localization in a hospital environment using Random Forest classifiers

• A system able to localize patients in a hospital environment is proposed.
• The system relies on RFID technology and a combination of Random Forest classifiers.
• The system has been deployed in the emergency unit of a large Italian hospital.
• Accuracy, precision, complexity, robustness and scalability have been evaluated.
• Patients localization is correctly performed in 98% of cases.

This paper proposes a new indoor localization system, based on RFID technology and a hierarchical structure of classifiers. This system has been specifically designed to work in unfriendly scenarios, where transmissions could be disturbed by other electronic devices or shielded walls. The infrastructure has been deployed and evaluated in the emergency unit of a large Italian hospital (48 rooms covering about 4000m2) to detect the room where lost or forgotten patients lie. Extensive experiments show the potential of such technology for indoor localization applications in terms of accuracy, precision, complexity, robustness and scalability. In 98% of cases the system localizes the correct room (83%) or one of its adjacency (15%).