Symbol Denoising in High Order M-QAM using Residual learning of Deep CNN
Saud Khan, Komal S. Khan, Soo Young Shin · 2019
This paper presents an integrating concept of de-noising convolutional neural networks (DnCNN) with quadrature amplitude modulation (QAM) for symbol denoising. DnCNN is used to estimate and denoise the Gaussian noise from the received constellation symbols of QAM with unknown noise level. Proposed system shows a significant gain in terms of peak signal-to-noise ratio, system throughput and bit-error rate; in comparison with conventional QAM systems. The basic concept, system level integration, and simulated performance gains are presented to elucidate the concept.