RIDNet Assisted cGAN Based Channel Estimation for One-Bit ADC mmWave MIMO Systems

Abstract

The estimation of millimeter wave (mmWave) massive multiple input multiple output (MIMO) channels becomes compelling when one-bit analog to digital converters (ADCs) are utilized. Furthermore, as the number of antenna increases, pilot overhead scales up to provide consistent channel estimation, eventually degrading spectral efficiency. This study presents a channel estimation approach that combines a conditional generative adversarial network (cGAN) with a novel blind denoising network with a sparse feature attention mechanism. Performance analysis and simulations show that using a cGAN fused with a feature attention-based denoising neural network significantly enhances the channel estimation performance while requiring less pilot transmission.