Information Geometry for Radar Target Detection with Total Jensen-Bregman Divergence

Area of Science:

• Radar Systems Engineering
• Information Geometry
• Signal Processing

Background:

• Traditional radar target detection methods face challenges with complex signal correlations.
• Modeling data correlations as Hermitian positive-definite (HPD) matrices offers a robust approach.
• Information geometry provides a powerful framework for analyzing matrix spaces.

Purpose of the Study:

• To propose a novel radar target detection algorithm leveraging information geometry.
• To introduce Jensen-Bregman divergences as distance metrics for HPD matrices.
• To develop a decision rule for target detection based on median matrices.

Main Methods:

• Modeling sample data correlations as Hermitian positive-definite (HPD) matrices.
• Utilizing total Jensen-Bregman divergences (square loss, log-determinant, von Neumann) as distance-like functions.
• Defining median matrices based on these divergences.
• Implementing a decision rule comparing divergences between reference and test cells.

Main Results:

• The proposed algorithm effectively utilizes information geometry for radar target detection.
• Jensen-Bregman divergences provide a robust measure of distance in the HPD matrix space.
• Median matrices derived from these divergences aid in decision making.
• Performance analysis demonstrates superior detection over conventional methods.

Conclusions:

• The proposed information geometry-based radar target detection algorithm offers enhanced performance.
• The use of Jensen-Bregman divergences and median matrices represents a significant advancement.
• The method shows promise for both simulated and real-world radar applications.