Fall Detection with Three-Axis Accelerometer and Magnetometer in a Smartphone

As an aging society approaches, the issue of injuries from falls by the elderly has emerged anew. Because of the dangers involved, research has recently been conducted on various types of falls; in particular, considerable research has been carried out on the detection and prevention of falls. We design an algorithm for the detection of falls using smartphones equipped with three-axis accelerometers and magnetometers. We propose a method of fall detection that recognizes a fall if the magnitudes of acceleration and angular displacement exceed given thresholds. Experiments to detect falls are performed in four directions: forward, backward, left, and right. Based on data from 200 experimental falls, obtained by fastening a smartphone to a belt worn around the waist, an overall detection rate of 95% was achieved, corresponding to direction-specific rates of 94% for forward falls, 100% for backward falls, 94% for leftward falls and 92% for rightward falls.