Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Novel Approach to WaveNet Architecture for RF Signal Separation with Learnable Dilation and Data Augmentation

Yu Tian, Ahmed Alhammadi, Abdullah Quran, Abubakar Sani Ali

TL;DR

This paper tackles co-channel RF signal separation in crowded spectra by using a modified WaveNet architecture with learnable dilation and padding to adapt its receptive field to complex RF mixtures. It pairs architecture refinements with a data augmentation strategy and a challenge-driven validation setup, training on $1100$ segments of $40960$-sample signals across $11$ SINR levels and optimizing via mean squared error. The results show a $58.82\%$ SINR improvement at BER $=10^{-3}$ for OFDM-QPSK with EMI Signal 1 and a $5$ dB SINR reduction (≈$33.33\%$ gain) for QPSK with Comm Signal 2, along with lower MSE, and the method achieved first place in the datadrivenrf2024 challenge. The work highlights the effectiveness of learnable dilation in WaveNet for RF separation and suggests practical applicability with potential for real-time adaptation and broader RF environments.

Abstract

In this paper, we address the intricate issue of RF signal separation by presenting a novel adaptation of the WaveNet architecture that introduces learnable dilation parameters, significantly enhancing signal separation in dense RF spectrums. Our focused architectural refinements and innovative data augmentation strategies have markedly improved the model's ability to discern complex signal sources. This paper details our comprehensive methodology, including the refined model architecture, data preparation techniques, and the strategic training strategy that have been pivotal to our success. The efficacy of our approach is evidenced by the substantial improvements recorded: a 58.82\% increase in SINR at a BER of $10^{-3}$ for OFDM-QPSK with EMI Signal 1, surpassing traditional benchmarks. Notably, our model achieved first place in the challenge \cite{datadrivenrf2024}, demonstrating its superior performance and establishing a new standard for machine learning applications within the RF communications domain.

A Novel Approach to WaveNet Architecture for RF Signal Separation with Learnable Dilation and Data Augmentation

TL;DR

This paper tackles co-channel RF signal separation in crowded spectra by using a modified WaveNet architecture with learnable dilation and padding to adapt its receptive field to complex RF mixtures. It pairs architecture refinements with a data augmentation strategy and a challenge-driven validation setup, training on segments of -sample signals across SINR levels and optimizing via mean squared error. The results show a SINR improvement at BER for OFDM-QPSK with EMI Signal 1 and a dB SINR reduction (≈ gain) for QPSK with Comm Signal 2, along with lower MSE, and the method achieved first place in the datadrivenrf2024 challenge. The work highlights the effectiveness of learnable dilation in WaveNet for RF separation and suggests practical applicability with potential for real-time adaptation and broader RF environments.

Abstract

In this paper, we address the intricate issue of RF signal separation by presenting a novel adaptation of the WaveNet architecture that introduces learnable dilation parameters, significantly enhancing signal separation in dense RF spectrums. Our focused architectural refinements and innovative data augmentation strategies have markedly improved the model's ability to discern complex signal sources. This paper details our comprehensive methodology, including the refined model architecture, data preparation techniques, and the strategic training strategy that have been pivotal to our success. The efficacy of our approach is evidenced by the substantial improvements recorded: a 58.82\% increase in SINR at a BER of for OFDM-QPSK with EMI Signal 1, surpassing traditional benchmarks. Notably, our model achieved first place in the challenge \cite{datadrivenrf2024}, demonstrating its superior performance and establishing a new standard for machine learning applications within the RF communications domain.
Paper Structure (7 sections, 3 figures)

This paper contains 7 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Methodology
  3. Model-Related Contributions
  4. Data-Related Contributions
  5. Data Preparation and Model Training
  6. Results
  7. Conclusion

Figures (3)

  • Figure 1: Modified Wavenet with Learnable Dilation and Padding
  • Figure 2: An Illustration of Learnable Dilation Rate
  • Figure 3: Comparative Performance Analysis of our Proposed Method Vs. Different Methods