A Power Spectrum Maps Estimation Algorithm Based on Generative Adversarial Networks for Underlay Cognitive Radio...

Area of Science:

• Wireless Communication
• Signal Processing
• Machine Learning

Background:

• Estimating power spectrum maps (PSMs) is crucial for detecting idle radio resources in underlay cognitive radio networks.
• Obtaining accurate radio propagation characteristics presents a significant challenge in existing methods.
• Traditional approaches often rely on imprecise or biased propagation assumptions.

Purpose of the Study:

• To propose a novel algorithm for accurate PSMs estimation using generative adversarial networks (GANs).
• To address the challenge of acquiring radio propagation characteristics for improved resource detection.

Main Methods:

• Developed a deep learning-based regression model for PSMs estimation.
• Transformed the estimation task into an image reconstruction problem using color mapping.
• Designed a Maps' Estimation GANs (MEGANs) framework with a generator and a discriminator.
• Trained the generator to extract propagation features and generate PSMs, while the discriminator refines the generator's output.

Main Results:

• The MEGANs algorithm continuously improved generator and discriminator performance through iterative training.
• Achieved high-accuracy PSMs estimation by learning accurate radio propagation features.
• Simulation results confirmed superior performance compared to conventional estimation methods.

Conclusions:

• The proposed MEGANs algorithm offers a more accurate and robust approach to PSMs estimation in cognitive radio networks.
• This method effectively overcomes the limitations of traditional techniques by learning propagation characteristics directly.
• Enhanced detection of idle radio resources is achievable through this advanced deep learning framework.