Evaluation of Multiple Bioradar-based Fall Detection System’s Performance on a Test Sample including Recordings with Pets, moving Appliances and Household Activity

The global trend towards the population aging makes the medical problems of the elderly, in particular falls, more significant from both moral and practical points of view. Machine learning models, the use of which is currently the main approach to the fall detection, are often trained and evaluated on data that differ from real conditions. In this paper, the performance of a previously developed radar-based fall detection system was evaluated on a test sample, which includes recordings with pet activity, appliance movements and household activity. Such movements presence a great issue because they can lead to false positive alarms.