Fall Detection Using UWB Data

Abstract

In this study, fall detection was aimed using only distance data from Ultra-Wideband (UWB) systems. After performing missing data imputation, noise reduction, and data standardization, One-Dimensional Convolutional Neural Network (1D-CNN), Long Short-Term Memory (LSTM), and Bidirectional Long Short-Term Memory (Bi-LSTM) models were trained, successfully performing high-accuracy fall detection. The models were tested on individuals not included in the training set, demonstrating successful generalization performance. The proposed method, which does not require fixed anchor position information or additional sensors, can be easily adapted to different environments and enables rapid intervention. © 2025 Elsevier B.V., All rights reserved.