DoA Estimation for FMCW Radar by 3D-CNN.

Area of Science:

• Electrical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Direction-of-arrival (DoA) estimation is crucial for radar systems.
• Traditional methods like MUSIC offer high resolution but have limitations.
• Deep learning presents a promising alternative for advanced DoA estimation.

Purpose of the Study:

• To propose and evaluate a novel 3D Convolutional Neural Network (3D-CNN) for DoA estimation in FMCW radar.
• To leverage deep learning for enhanced feature extraction and phase shift information capture.

Main Methods:

• Utilizing a 3D-CNN architecture to process FMCW radar data.
• Extracting spectrum features from the region of interest (RoI) in the radar data cube.
• Capturing spectrum phase shift information along the antenna axis for DoA determination.

Main Results:

• The proposed 3D-CNN method demonstrates superior performance compared to existing algorithms.
• Simulation and experimental results validate the effectiveness of the 3D-CNN approach.
• The study also identifies and discusses the limitations of the proposed method.

Conclusions:

• 3D-CNN is a highly effective deep learning approach for FMCW radar DoA estimation.
• The method successfully captures essential phase shift information for accurate DoA determination.
• Further research can build upon these findings to refine radar signal processing techniques.