Fall-detection system using mm-wave radar
Developed a mmWave radar system to detect human falls using doppler shift analysis and motion intensity tracking. Tuned radar parameters (chirp timing, bandwidth, and velocity resolution) for optimal performance in ceiling-mounted environments, achieving 8.33 cm range resolution and 9.26 m/s max velocity detection. Designed and 3D-printed ceiling tile analogs to evaluate signal attenuation through obstructions. Created a Matlab-based pipeline for processing radar cube data, using spatial filtering and temporal thresholds to identify falls with high accuracy and low false positive rates. Validated across multiple real-world scenarios including moving bystanders and seated motion.
Zoomed-in setup picture with 3D printed components and custom-sized blockers, mimicking the ceiling
Description
Detects and classifies falls in real time using 60 GHz mmWave radar, analyzing doppler shifts and motion intensity.
Designed for environments like hospitals, elderly care facilities, and grocery stores where rapid fall detection can improve safety.
Can operate behind ceiling tiles or visual barriers, leveraging radar’s material-penetrating properties for discreet installation.
Captures velocity and range data across 128 chirps/frame and a 1.8 GHz bandwidth, achieving 8.33 cm range resolution and 9.26 m/s max velocity.
Uses custom Matlab post-processing to identify falling patterns in radar data, comparing motion intensity across consecutive frames.
Fall detection is validated across multiple scenes and obstructions (0, 4, and 8 mm ceiling tiles), with high robustness and minimal false positives.
Learning Outcomes
1. Technical Skills & Knowledge
Radar fundamentals: doppler shift, range resolution, RCS, and radar equation
Worked with TI mmWave radar and its data capture tools
Designed and evaluated signal obstruction models using 3D-printed tiles
Optimized radar parameters (chirp duration, idle time, bandwidth, chirps/frame)
Wrote custom Matlab data pipelines for radarcube processing and fall classification
Performed frequency-domain analysis, FFT processing, and spatial filtering
2. General Project/Practical Skills
Experimental Design: Built physical test rigs to emulate real-world radar deployments
Performance Benchmarking: Evaluated under multiple test scenarios with varied occlusions
Error Analysis: Identified and explained false positives due to radar orientation artifacts
Documentation: Wrote structured reports, created visualizations, and published code + datasets
Future Planning: Proposed next steps including ceiling mounting, antenna arrays, and real-time angular tracking
Data and Visualization
Isolated Fall | Fall, moving people | No fall, moving people | Standing/Sitting | |
|---|---|---|---|---|
No wall | Fall Detected | Fall Detected | No Fall Detected | No Fall Detected |
4mm wall | Fall Detected | Fall Detected | Fall Detected | No Fall Detected |
8mm wall | Fall Detected | Fall Detected | No Fall Detected | No Fall Detected |
The original source files, as well as GIFs for the offset data can be found here. If viewing as a PDF, the original google doc including the code and equipment information can be found here for better viewing.
This project uses a 60 GHz radar sensor to detect human falls by analyzing doppler shifts and motion intensity over time. I optimized radar parameters for velocity tracking, designed realistic ceiling-tile barriers for testing, and built a robust Matlab post-processing pipeline that filters and classifies movement data. The system detects falls accurately in a variety of scenarios—even with obstructions—demonstrating a strong match for real-world use in hospitals or elderly care environments.
A mmWave radar system was developed to detect human falls by analyzing doppler shifts and motion intensity, achieving high accuracy and low false positives even with obstructions. Key features include optimized radar parameters for ceiling-mounted environments, a custom Matlab processing pipeline, and validation across various real-world scenarios. The project emphasizes technical skills in radar fundamentals, experimental design, and performance benchmarking, with future plans for enhancements like ceiling mounting and real-time tracking.