High-Resolution Image Synthesis Based Channel Estimation
Sachin Kumar Agrawal, Ayush Yadav, Owais Salim · 2023
This paper presents a novel approach for estimating the time-frequency response of a rapidly fading communication channel in an OFDM system. The technique employs deep learning methods for the estimation of unknown channel response matrix values, given known pilot site values. The proposed method utilizes a universal pipeline based on deep image processing methods, such as image super-resolution and image restoration. The pilot values are treated as a low-resolution image, and the channel is estimated using a super-resolution (SR) network and a denoising iterative residual reconstruction (IRR) network. The paper also includes a practical example of the proposed pipeline and demonstrates that it outperforms the previously used channel net approach. Furthermore, the proposed method achieves performance comparable to the minimum mean square error (MMSE) approach when all channel characteristics are known.