Activity recognition from interactions with objects using dynamic Bayesian network
知仁 猪股, Futoshi Naya, Noriaki Kuwahara, Fumio Hattori, Kiyoshi Kogure · 2009
A nursing activity recognition method from nurses' interactions with the tools and materials he/she touched has been developed for preventing the cause of medical accidents and incidents. The method detects an interaction between a nurse and a tool or material by using a RFID tag system. From interaction data, activities are recognized by using the Dynamic Bayesian Network (DBN) framework. This paper focuses on recognizing the twelve activity steps in the drip injection task. In an experiment, we obtained the 95.4% accuracy in recognizing these steps.