Video based vomit and facial foreign object detections for baby watch and safety

Nowadays, because of the low-birth rate and busy working parents, the baby's security is more and more important. The development of the intelligent baby watch and care technology can promote the baby's security in indoor and home environments. In this paper, we develop an eye features based mouth region detection method to detect abnormal conditions for baby watch. First, we use the proposed eye and cross-node filters to detect the eye zone and eye features. Second, according to the eye features positions, we find the correct mouth region by the geometric method, and detect dangerous conditions around the mouth, e.g. vomits. Simulation results show that the accuracy rate of the vomit and facial foreign object detections is up to 88%.