Fall Detection And Pre-impact Warning Based On MEMS Inertial Sensors

Fall monitoring systems aims at classifying falling behaviors (Fall) from activities of daily life (ADL), making it easy to take measures to prevent or warning falls. Fall monitoring systems include fall detection and fall pre-impact warning.Researching on fall monitoring systems has important social meaning under population aging. This paper studies the fall detection and pre-impact warning systems based on MEMS (Micro-electro-mechanical System) inertial sensors, the study consists of three parts.First, the article divided human activity patterns into ADL and fall, then set experiment patterns, organized and accomplished the fall experiments, with9inertial sensors worn in different body parts to record the kinematic and posture data of experimental object.Second, the article focused on the fall monitoring algorithms based on kinematic. The study found that in the impact phase of a fall event, set the acceleration threshold in21-28m/s2can get100%of the fall detection sensitivity and specificity, it is the best fall detection algorithm; in the pre-impact phase, set the acceleration threshold value in4.5m/s2, the detection sensitivity of the pre-impact warning algorithm can reaching98.61%and with a specificity of98.61%, the mean of the algorithm’s warning lead time is300ms; fall detection algorithm based on speed signal, set the speed threshold at-1m/s can get the detection sensitivity of98.61%under the premise of no false positives. The result of the experiments show that waist and chest are the most suitable place for emplace fall sensors.Third, the article conducted a further research on fall theory. At the base of divided a fall event into pre-fall phase, critical phase, post-fall phase and recovery fall in a real acceleration signal, then this paper divided critical phase into pre-impact phase, impact phase, post-impact phase according to the human body’s impact. Fall’s lead time is an important part of the fall theory. This study purposed a theoretical definition of fall lead time and found that the theoretical value of the lead time is about500±300ms.In summary, this paper systematically studied fall detection and pre-impact warning, and focused on fall monitoring algorithm based on kinematic characteristics. This paper, designed experiments, analyzed and processed data, obtained an ideal experimental result. On this basis, the paper continued in-depth research on fall theory, and improved it.