Design of smartwatch-assisted fall detection system via smartphone

Falling is an important health risk, especially for the elderly people. This situation prevents individuals from living independently. Automatic and high accuracy detection of the falls will contribute in preventing the negative situations that may occur. In this study, a mobile solution with a new architecture for the detection of falls is presented. For this purpose, motion sensor data has been collected simultaneously from smartwatch and smartphone which works with Android operating system. Data sets for both smartwatch and smartphone have been created by labeling the falls and actions which are not falling in the data. The performances of Random Tree, Bayesian Network and Logistic Regression methods have been tested on these datasets and the Logistic Regression has given the best result on smartwatch dataset and Bayesian Network method has given the best result on smartphone dataset.