A Novel Feature Set Extraction Based on Accelerometer Sensor Data for Improving the Fall Detection System

Because falls are the second leading cause of injury deaths, especially in elderly according to WHO statistics, there have been a lot studies on developing fall detection and warning system. Many approaches based wearable sensors, cameras, Infrared radar, etc., proposed detect efficiently. However, it still faces many challenges due noise no clear definition activities. This paper proposes new way extract 44 features time domain, frequency Hjorth parameters deal with this. The effect feature set has evaluated several classification algorithms, such as SVM, k-NN, ANN, J48, RF. Our method achieves relative high performance (F1-Score metric) detecting non-fall activities, i.e., 95.23% (falls), 99.11% (non-falls), 96.16% 99.90% (non-falls) for MobileAct 2.0 UP-Fall datasets, respectively.