Reconstruction of OFDM Signals Using a Dual Discriminator CGAN with BiLSTM and Transformer.

Area of Science:

• Electrical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Traditional Orthogonal Frequency Division Multiplexing (OFDM) signal reconstruction methods face challenges with complexity and suboptimal performance.
• Extracting intricate temporal information from OFDM signals using conventional Convolutional Neural Networks (CNNs) is difficult.

Purpose of the Study:

• To propose a novel dual discriminator Conditional Generative Adversarial Network (CGAN) model for improved OFDM signal reconstruction.
• To enhance the extraction of temporal and correlational information within OFDM signals for more accurate reconstruction.

Main Methods:

• A dual discriminator CGAN model is developed, integrating BiLSTM and Transformer networks.
• The first discriminator uses BiLSTM to capture temporal details from In-phase and Quadrature-phase (IQ) and Amplitude and Phase (AP) sequences.
• The second discriminator incorporates Vision Transformer (ViT) concepts, treating IQ sequences as images for enhanced correlation analysis.

Main Results:

• High-quality reconstruction of time series waveforms, constellation diagrams, and spectral diagrams for BPSK, QPSK, and 16QAM modulation formats.
• The proposed algorithm demonstrates superior signal quality compared to existing reconstruction methods.
• Effective management of computational complexity while achieving improved reconstruction performance.

Conclusions:

• The dual discriminator CGAN model effectively addresses limitations in traditional OFDM signal reconstruction.
• The integration of BiLSTM and Transformer architectures significantly enhances the model's ability to process complex signal data.
• This approach offers a promising solution for advanced communication signal reconstruction in countermeasures and signal processing applications.