A Smart Vision-based Human Fall Detection System for Telehealth Applications

S.-G. Miaou, F.-C. Shih, and C.-Y. Huang (Taiwan)


Human fall detection, omnidirectional camera, RFID, and personal information


The elderly and the people with certain health problems are more likely to experience fall accidents than other population groups. Some accidents are even life-threatening if proper care is not immediately available. This paper proposes a human fall detection system for telehealth applications. When a fall accident is detected, a call-for-help signal will be sent automatically and immediately. The proposed system is vision-based and it mainly includes a MapCam (omni-directional camera), image processing and recognition algorithms, and an RFID set. We use one MapCam to replace multiple traditional cameras because it can capture a 360 viewing angle of a surveillance space in single shot. Since only one camera is considered and a novel and powerful classification feature is extracted from MapCam images, the associated recognition algorithm for fall detection can be fairly simple. With the RFID, personal information (such as age and health history) can be adopted to customize the detection sensitivity for each individual in order to reduce the probability of false alarms for less likely events and put more attention on more likely events. The proposed system achieves about 91% successful fall detection rate according to the experimental results, where various walking paths and falling directions are tested.

Important Links:

Go Back