Human Activity Benchmark Classification Using Multilayer Artificial Neural Network

Human induced floor vibrations have recently been proposed to track human activity for a number of applications such as health care and surveillance. For example, floor vibrations can be used to identify human falls, or the existence of an intruder in a room. In these applications, the acceleration signals should be classified accurately to eliminate false positives. In this paper, a multi-layer artificial neural network is used to classify floor vibrations. Data from a previously published benchmark problem, which consists of seven types of human activities, is used to train and test the algorithm. Results show the capabilities of a multilayer artificial neural network in human activity classification.