An Analytical Comparison of Datasets of Real-World and Simulated Falls intended for the Evaluation of Wearable Fall Alerting Systems

The vast majority of works on fall detectors have been traditionally evaluated using ‘synthetic’ datasets, i.e. samples generated under laboratory conditions, in a controlled testbed in which a group of volunteers simulate o ‘mimic’ falls. This study tries to assess the adequacy of this testing procedure, as long as the dynamics of real falls may substantially differ from those exhibited by these simulations.