Channel Estimation for Massive MIMO Systems via Polarized Self-Attention-Aided Channel Estimation Neural Network

Area of Science:

• Wireless Communication
• Deep Learning
• Signal Processing

Background:

• Massive MIMO systems require accurate channel estimation for optimal performance.
• Traditional methods often suffer from low accuracy and high computational complexity.
• Deep learning (DL) offers a promising avenue for enhancing channel estimation.

Purpose of the Study:

• To propose a novel DL-assisted channel estimation algorithm for massive MIMO systems.
• To address the limitations of conventional channel estimation techniques.
• To introduce an efficient neural network architecture for improved performance.

Main Methods:

• The channel estimation problem is reframed as an image denoising task.
• A new polarized self-attention-aided channel estimation neural network (PACE-Net) is developed.
• A dedicated channel dataset is constructed for training and testing.

Main Results:

• PACE-Net demonstrates superior normalization mean square error (NMSE) performance.
• The proposed algorithm outperforms traditional and other DL-assisted methods.
• Significant reduction in computational complexity compared to MMSE estimation is achieved.

Conclusions:

• The proposed DL-assisted channel estimation algorithm offers enhanced accuracy and efficiency.
• PACE-Net provides a viable solution for complex channel estimation challenges in massive MIMO.
• The image denoising approach effectively tackles limitations of conventional methods.