Ultra-Wideband Radar-based Gait Analysis System for Non-Contact Knee Angle Monitoring

Area of Science:

• Biomechanics
• Sensor Technology
• Artificial Intelligence

Background:

• Clinical gait analysis relies on knee angle variations, typically measured by marker-based motion capture.
• Current methods face limitations such as being labor-intensive, confined to labs, and requiring specific environmental conditions (lighting, line-of-sight).
• Ultra-wideband (UWB) radar presents a promising alternative to overcome these constraints for gait analysis.

Purpose of the Study:

• To introduce a novel Ultra-wideband (UWB) radar-based system for capturing walking gait and estimating knee angles.
• To develop and validate a new detection intensity level algorithm (DILA) and a PointNet-inspired neural network (DIL-PointNet) for accurate knee angle measurement.
• To assess the system's potential for long-term, remote gait monitoring outside of controlled laboratory settings.

Main Methods:

• Development of a UWB radar system for gait data acquisition.
• Implementation of a novel Detection Intensity Level Algorithm (DILA) for processing radar signals.
• Integration of a PointNet-inspired neural network (DIL-PointNet) for 3D point cloud registration and knee angle estimation.
• Validation of the UWB system against a gold-standard marker-based motion capture system (Qualisys).

Main Results:

• The proposed UWB radar system successfully captured walking gait data.
• The DIL-PointNet model achieved accurate knee angle estimations.
• Validation against marker-based systems yielded a Mean Absolute Error (MAE) of 7.9° and a Root Mean Squared Error (RMSE) of 11.1°.
• The system demonstrated effectiveness in registering detections in 3D space.

Conclusions:

• The UWB radar-based gait analysis system provides a viable alternative to traditional methods.
• The developed DILA and DIL-PointNet algorithm show significant potential for accurate knee angle estimation.
• This technology enables remote and long-term gait monitoring in real-world environments, overcoming previous limitations.