Enhanced characterization of an accelerometer-based fall detection algorithm using a repository
Kuang-Hsuan Chen, Yu-Wei Hsu, Jing‐Jung Yang, Fu‐Shan Jaw · Instrumentation Science & Technology · 2016
Falling is a common accident that can lead to serious injury among the elderly. To reduce injuries resulting from falls, automatic detection allows immediate medical assistance. Hence, various fall detectors, including body sensors and smartphones, have been developed in recent decades. In our previous study, an accelerometer-based fall detector was designed with high accuracy for simulated falls performed by young volunteers. However, there are significant differences between the acceleration signals generated by simulated and real falls. Simulated devices do not accurately assess the sensitivity and specificity of falls. Hence, the goal of this study is to access the accuracy of our designed accelerometer-based fall detection algorithm using a real-world repository. The results showed that the algorithm accurately characterizes real falls. Differences between our approach and previously published algorithms are discussed. This study is expected to assist in the design of more effective practical fall detection algorithms.