Doppler-Spectrum Feature-Based Human-Vehicle Classification Scheme Using Machine Learning for an FMCW Radar Sensor.

Area of Science:

• Radar Signal Processing
• Machine Learning for Sensor Fusion
• Automotive and Human Detection Systems

Background:

• Frequency-modulated continuous wave (FMCW) radar is crucial for object detection.
• Distinguishing between humans and vehicles based on radar signatures presents challenges.
• Existing methods may lack the precision required for reliable classification.

Purpose of the Study:

• To develop a novel human-vehicle classification scheme using Doppler-spectrum features for FMCW radar.
• To introduce and validate three new features: scattering point count, scattering point difference, and magnitude difference rate.
• To assess the performance of machine learning classifiers (SVM and BDT) with these features.

Main Methods:

• Utilized a 24-GHz FMCW radar sensor and real-time data acquisition.
• Extracted three novel Doppler-spectrum features from successive frames of walking humans and moving vehicles.
• Trained and validated Support Vector Machine (SVM) and Binary Decision Tree (BDT) classifiers using the extracted features.

Main Results:

• The proposed Doppler-spectrum feature-based classification scheme demonstrated high efficacy.
• Achieved an average classification accuracy exceeding 99% for walking humans.
• Achieved an average classification accuracy exceeding 96% for moving vehicles.

Conclusions:

• The novel Doppler-spectrum features are effective for distinguishing between humans and vehicles.
• The proposed classification scheme offers a robust and accurate solution for FMCW radar applications.
• The high performance validates the potential of this approach in real-world scenarios.