Fall Detection on Humans Using Threshold Method Base On Smartphone Accelerometer Data

by Mardi Hardjianto, Jazi Eko Istiyanto, Subanar, Agfianto Eko Putra

Abstract – Many fall detection systems are developed using accelerometer and gyroscope on smartphone. In existing fall detection systems, smartphone location placement is generally selected from the beginning and cannot be repositioned. The movement of smartphone decreases the accuracy of the system. This research provides fall detection model using accelerometer
on smartphone and the placement of the smartphone could vary as of six predetermined positions. The method used for fall detection is threshold method applying only one parameter, the value of resultant acceleration. There are two thresholds used in this research, upper threshold and lower threshold. The testing of fall detection model is conducted by eight volunteers (four males and four females), and resulted in 98.1% of accuracy, 96.9% of sensitivity, and 100 specificity.

source: IJASCSE