A Test-Bed Based Evaluation of Pervasive Fall Detection Schemes

Fall has been a major cause of death for the people above the age of 65. In most cases these deaths can be avoided if the event of fall gets reported in time to the health care personals. This work selects three well know accelerometer based fall detection algorithms from literature and analyze their performance based upon Sensitivity, Specificity and Accuracy. It has been observed that the efficiency of a fall detection algorithm can be greatly improved by following a two step process. First detect all types of event, fall and ADL (activities of daily life) based on observed resultant acceleration (RSS) value. Second, classify observed event in fall or ADL through detailed analysis of recorded RSS readings.