Study of the Detection of Falls Using the SVM Algorithm, Different Datasets of Movements and ANOVA

Abstract

Falls are becoming a major public health problem, which is intensified by the aging of the population. Falls are one of the main causes of death among the elderly and in population groups that develop risk activities. In this sense, technologies can provide solutions to improve this situation. In this work we have analyzed different repositories of movements and falls designed to test decision algorithms in automatic fall detection systems. The objectives of the study are: firstly, to clarify what are the characteristics of the most significant accelerometry signals to identify a fall and secondly, to analyze the possibility of extrapolating the learning achieved with a certain database when tested with another one. As a novelty with respect to other works in the literature, the statistical significance of the results has been systematically evaluated by the analysis of variance (ANOVA).