Comparison of Data-Processing Procedures for Fall Detection in Impulse-Radar-Based Systems for Monitoring Human Activity

The classification of human activities, based on impulse-radar data, is a burgeoning field within biomedical signal processing, showing potential benefits for healthcare monitoring and rehabilitation services. This article contains the results of a comparative study of 63 data-processing procedures designed for the recognition of human activity. The study has been based on a rich set of impulse-radar data representative of five human actions: walking, sitting, lying down, crouching, and falling down. The criteria of comparison have been selected in such a way as to put a special emphasis on the latter action, which is of critical importance in the healthcare monitoring of elderly or disabled persons. The ranking of the procedures, resulting from the comparative study, enabled the authors to select two of them for further development aimed at their application in a monitoring system, whose functionality includes both the systematic recording of the activity of monitored persons and the reliable detection of their falls.