Transformer-Based Nonlinear Blind Source Separation for Anti-Jamming in DSSS Satellite Communications.

Area of Science:

• Electrical Engineering
• Signal Processing
• Satellite Communications

Background:

• High-power jamming can saturate satellite receiver front ends, causing nonlinearity.
• Conventional anti-jamming methods fail due to the violation of linear superposition assumptions.
• Nonlinear interference degrades satellite signal quality and receiver performance.

Purpose of the Study:

• To develop an advanced anti-jamming technique for direct-sequence spread-spectrum (DSSS) satellite receivers operating in nonlinear regimes.
• To address the challenge of separating desired DSSS signals from strong jamming interference in nonlinear environments.
• To enhance the robustness and performance of satellite receivers against sophisticated jamming attacks.

Main Methods:

• Formulated dual-receiver DSSS anti-jamming as a nonlinear source-separation problem using stacked I/Q observations.
• Proposed a time-domain separator combining a convolutional front end and a Transformer encoder.
• Employed supervised pretraining on synthetic nonlinear mixtures and an optional mixture-only adaptation step for refinement.

Main Results:

• The proposed method significantly improved the scale-invariant signal-to-noise ratio (SI-SNR) from -4.79 dB to 17.73 dB.
• Failure rates (SI-SNR < 0 dB) decreased dramatically from 60.7% to 2.3% across diverse test conditions.
• Enhanced robustness under high jammer-to-signal-ratio (JSR) conditions through oversampling and JSR-aware loss reweighting.

Conclusions:

• The developed nonlinear source-separation technique effectively combats high-power jamming in satellite receivers.
• The method offers a robust solution for improving DSSS signal recovery in challenging nonlinear RF environments.
• Fine-tuning with nonlinear reconstruction consistency and communication priors further boosts performance in blind adaptation scenarios.