A Real-time Human Tracking System with Fall Detection Capability

Fall is an unexpected phenomenon which occurs frequently in the elderly and is a major source of expenditure in healthcare systems. Studies show that more than 30% of the elderly people over the age of 65 experiences falls more than once a year. One of the major consequences of falls is 'long-lie' which refers to the state of remaining lied down on the ground after a fall for over an hour. 'Long-lie' as a result of losing consciousness and serious injuries might lead to death. As such, early detection of a fall is important, i. e. minimizing the time between fall time and rescue time. This research proposes a real-time system which is able to track a subject and activate an alarm for immediate medical assistance in the case of falling. The proposed system uses three accelerometers and three gyroscopes to detect a fall based on a set of thresholds, considering that falls have high peak values in accelerations, angular velocities, and tilts in short periods. This system is also able to distinguish between fall and non-fall activities and avoid false alarms. In this research, transmitting fall alarms and local position information of the subject to a monitoring room is achieved through a wireless network. Choosing a proper wireless network protocol for data transmission is a critical issue. In this application, ZigBee wireless standard is deployed due to its low power consumption, reliability, simple connectivity, mesh networking architecture, and desirable area of coverage. An active RF transceiver is attached to the subject along with the fall detection sensors to transmit an alarm signal in the case of a fall. In addition to the RF transceiver on the mobile subject, several RF transceivers are placed in designated areas of rooms and corridors of the building to create the beacon infrastructure of the wireless network. Moreover, the wireless network is configured in a way to pinpoint the location of the monitored subject in the building and transmits the position information to the monitoring room. Experimental results show that the proposed system has a local positioning accuracy within a 3m radius. Furthermore, highly reliable results have been achieved for fall detection. The usage of a wireless network based on the ZigBee protocol for detecting room level location and transmitting fall detection signals is a novel approach that has been investigated in this paper. Transmitting these critical data will result in timely reaction which can save fallers that would otherwise be found in severe conditions or even deceased.