Fall classification based on sensor data from smartphone and smartwatch

Abstract

A fall is an unintentional event in which a person comes to rest on the ground/floor/a lower level. Falls are the second driving cause of inadvertent injury deaths around the world. Falls pose a major health risk to children and the elderly (especially above the age of 65). Injuries from falls make it difficult for an individual to move around, do regular activities, or live on their own. Hence, studies on methods able to recognize Activities of Daily Living (ADLs) and to identify falls is on the rise. The acknowledgment of ADLs may permit to gather the sum of physical activity that a subject performs daily, whilst the timely identification of falls may aid in diminishing their consequences. To improve the accuracy of the classifier, the data from smartwatches and smartphones are combined. There are not many publicly accessible datasets incorporating data from both smartwatchesand smartphones. Hence, the data will be collected independently. The Decision tree (J48) classifier will be used to classify the falls. The evaluation metrics include: Precision, Recall, F1 Score and the confusion matrix.