Feature Extraction from Double Task Timed Up and Go Signals Applied to the Identification of Elderly Fallers.

Devices and sensors for identification of fallers can be used to implement actions to prevent these falls and to allow the elderly to live an independent life while reducing the long-term care costs. In previous studies the feature extraction is often dependent on step detection, the collected datasets are often unavailable, and a large number of features are attempted in order to sufficiently represent the faller and non-faller classes. In this study we present results of feature extraction on accelerometer data with the aim of identifying fallers among a group of health elderly. The features are extracted from three different Timed Up and Go (TUG) tests combined with simultaneous manual and cognitive tasks. As a result, a set of features is described to discriminate fallers from non-fallers based on the segmentation of the TUG trials, in particular using TUG with cognitive task, while conventional functional tests fail to do so. In addition, we show that the features can be generalised for unseen subjects, demonstrating the relevance of the study. We concluded that features extracted from segmented TUG trials acquired with simultaneous tasks has potential to improve performance when identifying fallers via accelerometer sensors.