IR-UWB Sensor Based Fall Detection Method Using CNN Algorithm

Area of Science:

• Gerontology
• Biomedical Engineering
• Computer Science

Background:

• Falls are a leading cause of fatal injuries and death in the elderly, necessitating effective detection methods.
• Existing fall detection systems often struggle to balance privacy, user convenience, and detection accuracy.
• Developing a system that addresses these competing demands is crucial for elderly care.

Purpose of the Study:

• To create a practical fall detection framework capable of classifying behaviors as "Fall" or "Activities of Daily Living (ADL)".
• To ensure the system preserves user privacy and maintains user convenience.
• To achieve high detection accuracy for falls in elderly individuals.

Main Methods:

• Utilized non-contact, unobtrusive impulse-radio ultra wideband (IR-UWB) radar to collect motion signal data.
• Applied a convolutional neural network (CNN) algorithm to process IR-UWB data for behavior classification.
• Collected data through real-world performance of various daily activities, including simulated falls.

Main Results:

• The developed CNN classifier demonstrated satisfactory performance in distinguishing between falls and ADLs.
• The combined IR-UWB radar and CNN approach proved effective in fall detection.
• The system achieved a notable level of detection accuracy while upholding privacy and convenience.

Conclusions:

• The study successfully demonstrated the feasibility of a practical fall detection system using IR-UWB radar and CNN.
• This integrated approach offers a promising solution for elderly fall monitoring, enhancing safety without compromising privacy or convenience.
• The findings support the development of advanced, user-centric fall detection technologies.