Smartphone based fall detection system

This paper describes the design of a smartphone based fall detection system and characterizes the preliminary efficacy of the proposed system in activities of daily living (ADLs). Using the embedded sensors available in a smartphone (i.e., accelerometer, gyroscope and magnetometer), kinematic analysis of movement can be performed in real-time, allowing for continuous monitoring of fall status. Fall sensing thresholds are defined based on angular rate of change (TH1), maximum acceleration (TH2), and maximum attitude change (TH3). TH1 is measured from the resultant pitch and roll angular velocity vector and defined as 3.1 rad/s (~180°/s). TH2 is measured from the resultant acceleration vector and defined as 1.6 g. TH3 is measured from the resultant vector of the pitch and roll angles, and defined at 0.59 rad (39°). A proof-of-concept study was performed on five ADL tasks: 1) comfortable walking, 2) stand-to-seated posture, 3) seated-to-standing posture, 4) pivoting at the waist to pick up an object, and 5) stand-to-seated-to-laying transition. No trials violated the defined thresholds for fall detection, signifying no false positives. These results are important for the definition of machine learning algorithms, currently under development, to minimize false positive and false negative fall detection events.