Related Experiment Video
Updated: May 28, 2026
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
Simulation-to-Real Trip-Fall Detection with Continuous-Wave Doppler Radar via Physics-Informed Kinematic Modeling and
1Department of Data Science for Business Innovation, Chuo University, Tokyo 112-8551, Japan.
Sensors (Basel, Switzerland)
|May 27, 2026
Summary
This study demonstrates that physics-informed simulation can effectively train radar systems for detecting falls in older adults, reducing the need for extensive real-world fall data. This approach enhances fall detection capabilities for public health.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Gerontology
Background:
- Falls in older adults represent a significant public health challenge.
- Acquiring large-scale real-world fall data for radar-based detection is ethically and practically challenging.
- Existing fall detection methods often require extensive real-world data for training.
Purpose of the Study:
- To evaluate the feasibility of using simulated data for training radar-based trip-fall detection systems.
- To develop a method for generating realistic synthetic radar signals for fall events.
- To assess the performance of a classifier trained on simulated data for distinguishing between falls, walking, and breathing.
Main Methods:
- Coupled a physics-informed kinematic trip-fall model with a continuous-wave (CW) Doppler radar observation model to synthesize I/Q signals and Doppler spectrograms.
- Employed domain randomization to vary parameters such as body size, fall direction, and sensor placement.
- Generated synthetic data for three classes: trip fall, walking, and seated quiet breathing.
Main Results:
- Simulated spectrograms accurately reproduced key time-frequency characteristics of measured enacted trip-fall signals (mean SSIM of 0.782).
- A ResNet-18 classifier trained solely on simulated data achieved a macro-F1 score of 0.912 on measured data, outperforming a real-data-trained baseline (0.748).
- The simulation-based approach significantly improved classification performance compared to the baseline (p=0.006).
Conclusions:
- Physics-informed simulation combined with domain randomization can effectively reduce the reliance on real trip-fall data, especially under limited-data conditions.
- This approach shows promise for developing robust radar-based fall detection systems.
- Further research is needed to establish robustness across diverse fall types, activities, environments, and radar systems.
