# Privacy-aware IoT Fall Detection Services For Aging in Place

arXiv: https://arxiv.org/abs/2506.22462 · HTML: https://mlsanity.com/paper/2506.22462/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 40 references that are in the map.
- First posted on arXiv 18 Jun 2025.

## Builds on (most cited first, 8 of 8)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Class-Balanced Loss Based on Effective Number of Samples](https://mlsanity.com/paper/1901.05555/) (2019)
- [InceptionTime: Finding AlexNet for Time Series Classification](https://mlsanity.com/paper/1909.04939/) (2019)
- [Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline](https://mlsanity.com/paper/1611.06455/) (2016)
- [An Empirical Survey of Data Augmentation for Time Series Classification with Neural Networks](https://mlsanity.com/paper/2007.15951/) (2020)
- [TTS-GAN: A Transformer-based Time-Series Generative Adversarial Network](https://mlsanity.com/paper/2202.02691/) (2022)
- [Data augmentation using synthetic data for time series classification with deep residual networks](https://mlsanity.com/paper/1808.02455/) (2018)
- [Dynamic Conflict Resolution of IoT Services in Smart Homes](https://mlsanity.com/paper/2110.07083/) (2021)

## Abstract (arXiv)

Fall detection is critical to support the growing elderly population, projected to reach 2.1 billion by 2050. However, existing methods often face data scarcity challenges or compromise privacy. We propose a novel IoT-based Fall Detection as a Service (FDaaS) framework to assist the elderly in living independently and safely by accurately detecting falls. We design a service-oriented architecture that leverages Ultra-wideband (UWB) radar sensors as an IoT health-sensing service, ensuring privacy and minimal intrusion. We address the challenges of data scarcity by utilizing a Fall Detection Generative Pre-trained Transformer (FD-GPT) that uses augmentation techniques. We developed a protocol to collect a comprehensive dataset of the elderly daily activities and fall events. This resulted in a real dataset that carefully mimics the elderly's routine. We rigorously evaluate and compare various models using this dataset. Experimental results show our approach achieves 90.72% accuracy and 89.33% precision in distinguishing between fall events and regular activities of daily living.