BROWSE
Detail
Journal Article
Detection of important features and comparison of datasets for fall detection based on wrist-wearable devices
Cited 12 time in 
Share 
Share
- Authors
- Jeong-Kyun Kim, Kangbok Lee, Sang Gi Hong
- Issue Date
- 2023-12
- Citation
- Expert Systems with Applications, v.234, pp.1-9
- ISSN
- 0957-4174
- Publisher
- Elsevier Ltd.
- Language
- English
- Type
- Journal Article
- Abstract
- Falls are a major cause of unintended injuries and death worldwide. Fall detection has received considerable attention in wrist-wearable devices such as smartwatches and bands. In the case of network congestion, such as during a disaster, the available transmission bandwidth may become limited. The device must send minimal data to the server to detect falls. We proposed a high-accuracy fall detection method using limited data by analyzing three open datasets acquired from a wrist sensor using artificial intelligence. The proposed model achieved a higher fall prediction accuracy than those of previous studies with each dataset. Moreover, by adopting an explainable artificial intelligence algorithm, the amount of data was reduced fourfold; however, the performance remained unchanged at ±1%. To verify the robustness of the model, they were validated against each other and within each dataset, and fifteen features were selected for fall detection. The proposed model obtained a high fall detection accuracy with less data than those used in previous studies, demonstrating the effectiveness of wrist-wearable devices.
- KSP Keywords
- Detection Method, Detection accuracy, Fall Detection, Fall prediction, High accuracy, Intelligence algorithm, Limited data, Prediction accuracy, Proposed model, Wearable Devices, artificial intelligence