Next-Generation Noise-Resilient Communication Receiver Design Using Denoising Autoencoders

Autoencoders have become a prominent focus in unsupervised learning research due to their ability to capture essential data features, perform efficient dimensionality reduction and aid in noise reduction. In this paper, we propose a novel approach that integrates autoencoders with deep learning classifiers for the efficient reception of Binary Phase Shift Keying signals. Specifically, three distinct autoencoders–Linear, Long Short-Term Memory, and Convolutional– are cascaded with deep learning classifiers to denoise received signals corrupted by Additive White Gaussian Noise. This streamlined methodology not only enhances signal quality and interpretability but also facilitates the development of a more efficient receiver, outperforming conventional designs that rely on multiple processing blocks paving the way for robust and adaptive communication systems.