Comparative Analysis of CNN Models for SNR Estimation
Abdullah Al Mahbub, Ijaz Ahmad, Seokjoo Shin · 2025
Signal-to-Noise Ratio (SNR) estimation is a fundamental task in wireless communication, directly influencing channel ability, energy efficiency, and link quality. Recent advancements in deep learning, particularly Convolutional Neural Networks (CNNs), have shown superior performance in SNR estimation, offering higher accuracy and adaptability to complex noise conditions. However, a key challenge is that CNN models may show reduced generalization when the training and testing data distributions are mismatched. Therefore, this paper presents an extensive performance analysis of multiple CNN models for SNR estimation, assessing their robustness in both matched and mismatched conditions. The study aims to find architectures that excel in feature extraction and keep reliability across varying noise distributions for cellular and satellite range of SNR. Experimental results highlight the strengths and limitations of different models, offering valuable insights into SNR estimation applications in wireless communication.