Channel Estimation with DnCNN in Massive MISO LEO Satellite Systems.

In low Earth orbit (LEO) satellite communication systems, obtaining accurate channel state information (CSI) is crucial for achieving high performance. Least squares (LS) channel estimation is a simple conventional channel estimation scheme, but it does not account for compensating for channel estimation errors. In this paper, we propose a channel estimation scheme with a machine learning-based denoising network for massive multiple-input single-output LEO satellite communication systems. Our proposed scheme uses a denoising convolutional neural network to reduce channel estimation errors from the LS estimator. The numerical results demonstrate that our proposed machine learning-based denoising network effectively improves the accuracy of the estimated channel from the LS channel estimator.