Deep Learning-Based Residual and Recovery Framework for Channel Estimation in 5G/B5G Networks
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Channel estimation in wireless communication has gotten progressively more complex with the growth of modern methods. This paper uses a deep learning framework using a multilayer perceptron (MLP) to estimate channels in fifth-generation (5G), beyond 5G (B5G), and sixth-generation (6G) millimeter-wave (mmWave) environments to improve the reliability and accuracy of signal transmission. As the process deals with communication in 6G environment, a synthetic data set generation module has been designed using a sparse multi-path propagation model for simulating 60 GHz mmWave channels in different signal-to-noise ratio (SNR) conditions. This paper investigates fast learning-based channel estimation for 6G mmWave systems. The framework introduced is a switchable two-stage MLP-based estimator, where a lightweight primary network performs residual correction over Least Square (LS) estimates and a secondary refinement network is selectively activated under challenging channel conditions. In order to avoid redundancy in learning across stages, a controlled corruption-recovery mechanism is used, which forces the secondary network to learn robust channel recovery rather than identity mapping. Simulation results show improved robustness of estimation in noise-limited regimes while preserving high-SNR performance with reduced unnecessary inference complexity.
Fields of Study & Subjects
Millimeter-Wave Propagation and Modeling
Advanced Wireless Communication Techniques
Advanced Wireless Communication Technologies
Publication details
- Journal / source
- International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
- Pages
- 1–6
- Publication date
- 26 February 2026
- Language
- EN