Infrared-Based Measurement and Lightweight Model Design for Non-Contact Fall Detection in Elderly Care

With the accelerating global aging population, the health risks posed by falls among the elderly have become a pressing social issue [1]. Meanwhile, achieving accurate, reliable, and ethically sound measurement of fall events has emerged as a critical challenge in the field of measurement. Some traditional wearable fall detection devices, while capable of monitoring movement status, have inherent drawbacks regarding measurement reliability: specifically, certain direct skin-contact models (e.g., closely worn flexible electrodes and biosensors) may be affected by sweat erosion, compromising stability and leading to measurement errors or false alarms. Moreover, their dependence on battery power can disrupt the continuity of motion parameter monitoring—critical for capturing the transient dynamics of falls. Meanwhile, vision-based solutions, though enabling non-contact measurement, pose ethical concerns as they collect privacy-sensitive visual information, which diminishes the acceptability of such measurement systems in practical elderly care scenarios.