An evaluation on threshold based fall detection

This abstract expands upon fall detection paradigms to form a more concrete understanding of what is considered a fall. Data was gathered by a waist mounted inertia sensor, consisting of a tri-axis gyroscope and tri-axis accelerometer, from falling events, Full Motion Falls (FMF) and Partial Motion Falls (PMF), and non-falling events, Active Daily Life (ADL) events and Recovered Falls (RF). These falls generated threshold values for the following parameters: acceleration, angular velocity, angular acceleration and total angle change. Our algorithm considered the event to be a fall if three of the four parameters exceed their threshold values. The system found true positives in 92% of cases, and true negatives in 84% of cases.