Improving Channel Estimation Through Gold Sequences

TL;DR

The paper tackles channel estimation in downlink NOMA under pilot contamination by proposing Gold sequence-based user separation and a Channel Prediction Function (CPF) that leverages fractional power allocation and partially decoded data. It incorporates data-aided estimation and a DL-based channel predictor to outperform traditional pilot-only methods. Results show Gold-coded NOMA can achieve lower SER than C-V-BLAST, with DL models enhancing channel predictions, though scalability becomes challenging as network size grows, motivating adaptive sequence strategies. Overall, the approach integrates coding, power control, and learning to improve robustness and performance in practical NOMA deployments.

Abstract

This study evaluates Non-Orthogonal Multiple Access (NOMA) systems using Gold coding and Conventional-V-BLAST (C-V-BLAST). Superimposed signals on shared subcarriers make NOMA user separation difficult, unlike MIMO. Gold sequences' orthogonal features may enhance user separation and channel estimation. A novel channel estimation approach uses fractional power allocation and partially decoded data symbols. A realistic simulation environment was created using AWGN, Rayleigh fading, and shadowing. Using pilot signals, power allocation, and data symbols, our Channel Prediction Function (CPF) surpasses pilot-based techniques.

Figures (7)

• Figure 1: Transmission block structure at the BS in NOMA network.
• Figure 2: SER performance comparison of a two-user NOMA system employing Gold sequence lengths 31, 63, and 127.
• Figure 3: SER performance comparison of Gold coding and C-V BLAST for a two-user NOMA system.
• Figure 4: A rolling window approach with a 2-minute window size was applied to the 10-minute dataset.
• Figure 5: Loss Function of Training and Validation Dataset.