Performances of an accelerometric device for detecting fall and loss of consciousness

Falls and loss of consciousness (FLoC) is the leading cause of serious health problems, above all in the elderly population, since the subjects involved are not able to ask for help and may lie in critical conditions for a long time, thus deteriorating into severe conditions which can even lead to death. Therefore, developing a device capable of automatically detecting a FLoC and activating an alarm call seems to be of utmost importance. This study intended to develop such a device using an accelerometer sensor. Four hundred and sixty simulated falls were performed by 20 subjects: 10 young subjects and 10 elderly subjects. The young subjects were asked to perform 200 FLoCs as well as 60 non common activities (NCAs), whereas the elderly subjects were asked to carry out only 200 activities of daily living (ADL). The signal used to detect the fall event was acquired by a single accelerometer placed on the subjects' belts. The test set was divided into two groups of the same size: Training Set (TS) and Verification Set (VS). The first set was meant to determine the related algorithm, whereas the second set was intended to check its reliability. The proposed algorithm was devised to detect the effects of the three phases of a FLoC (impact of the body on the ground, lying position and immobility) into the acceleration and jerk signals along the cranio-caudal axis (CCA). The correct detection of all FLoC cases and the absence of false positives among ADL corroborate the usefulness of the device proposed.