Online Testing in Machine Learning Approach for Fall Detection
Journal
2020 International Joint Conference on Neural Networks (IJCNN)
Date Issued
2020
Author(s)
Nuñez-Martínez, José
Pacheco, Sofia
Type
text::conference output::conference proceedings::conference paper
Abstract
Robust fall detectors are needed to reduce the time in which a person can receive medical assistance, and mitigate negative effects when a fall occurs. Robustness in fall detection systems is difficult to achieve given that there are still many challenges regarding performance in real conditions. Fall detection systems based on smartphones present good results following a traditional methodology of collecting data, training and evaluating classification models using the same sensors and subjects, yet fail to experiment and succeed in different realistic conditions. In this paper, we propose a methodology to build a solution for fall detection, and online testing changing the sensors and subjects of evaluation in order to provide a more flexible and portable fall detector. © 2020 IEEE.