Efficient Support Vector Regression for Wideband DOA Estimation Using a Genetic Algorithm.

Area of Science:

• Signal Processing
• Machine Learning
• Array Signal Processing

Background:

• High-precision direction of arrival (DOA) estimation is crucial for radar and communication systems.
• Existing methods often face challenges with wideband signals and computational complexity.

Purpose of the Study:

• To develop an efficient and high-performance wideband DOA estimation algorithm.
• To reduce computational load and storage requirements for resource-constrained scenarios.

Main Methods:

• Proposed an efficient support vector regression (SVR) architecture optimized by a genetic algorithm (GA).
• Utilized the two-sided correlation transformation (TCT) algorithm for efficient network training using reference frequency data.
• Implemented a preprocessing step to reduce the dimensionality of the array covariance matrix, leveraging its conjugate symmetry and elemental characteristics.

Main Results:

• Achieved high estimation performance and generalization capabilities for wideband DOA.
• Significantly reduced training time and system storage capacity by maintaining constant input feature dimensionality regardless of signal bandwidth.
• Demonstrated superior efficiency and performance compared to existing methods through experimental validation.

Conclusions:

• The proposed GA-optimized SVR method offers an efficient and effective solution for wideband DOA estimation.
• The dimensionality reduction technique is particularly beneficial for broadband and ultra-broadband signals in resource-constrained applications.
• The algorithm shows significant advantages in terms of performance, training efficiency, and storage requirements.