Blind Non-Data-Aided Signal-to-Noise Ratio Estimation with Convolutional Neural Networks

Signal-to-noise ratio (SNR) estimation has always been an important part of the communication systems and multiple methods have been developed within that scope. In this paper, an SNR estimation method, which can work under difficult conditions, blindness to the modulation type, that many other previously developed methods would fail under, is presented. In the proposed method, deep learning and convolutional neural networks (CNN), which have shown high performances in recent years, were exploited. In conclusion, deep learning methods, that have applications in numerous areas, and SNR estimation, which is an essential problem in the communication systems, were brought together, and experimental results were presented.