eHomeSeniors Dataset: An Infrared Thermal Sensor Dataset for Automatic Fall Detection Research

Area of Science:

• Gerontology
• Biomedical Engineering
• Computer Science

Background:

• Automatic fall detection (AFD) is crucial for elderly care, aiming for rapid detection to mitigate health consequences.
• Existing AFD systems often rely on wearable devices or cameras, posing privacy concerns or user dependence issues.
• Current public datasets lack realism, using young individuals in lab settings, which doesn't reflect older adults' fall characteristics.

Purpose of the Study:

• To introduce the eHomeSeniors dataset, a novel public dataset designed to address limitations in current fall detection research.
• To provide a more realistic and privacy-preserving dataset for developing and evaluating AFD systems for the elderly.
• To improve the accuracy and applicability of AFD systems by incorporating data that better represents real-world fall scenarios.

Main Methods:

• The eHomeSeniors dataset was collected using two types of privacy-friendly infrared thermal sensors.
• Data was gathered from young volunteers and performing artists guided by a physiotherapist to simulate realistic falls of older adults.
• Specific fall types were carefully selected based on a comprehensive literature review.

Main Results:

• The dataset offers a unique combination of privacy-friendly sensing and realistic fall emulation.
• It captures data reflecting the distinct falling characteristics of older adults, unlike existing datasets.
• The methodology allows for the creation of more robust and generalizable AFD algorithms.

Conclusions:

• The eHomeSeniors dataset represents a significant advancement for research in automatic fall detection for elderly care.
• Its realistic and privacy-conscious approach is expected to accelerate the development of effective commercial AFD systems.
• This dataset will enable the training of machine learning models that are better suited to the complexities of real-world elderly falls.