Self-Supervised Noise2Noise-Enhanced Denoising for Continuous-Scan Air-Plasma THz Spectroscopy

2026-08-17

AI Summary

This study addresses the limitations of low signal-to-noise ratio and prolonged averaging times in single-shot continuous-wave terahertz time-domain spectroscopy by proposing a self-supervised denoising method based on the Noise2Noise strategy. Leveraging a one-dimensional residual U-Net, the approach learns directly from noisy data pairs without clean labels, effectively suppressing systematic bias. Experimental results demonstrate that the denoised single-scan signal achieves quality equivalent to approximately 5.4 averaged raw scans, significantly outperforming conventional Wiener filtering. Crucially, this software-only solution enhances both reconstruction accuracy and measurement speed without requiring hardware modifications, establishing an efficient new paradigm for rapid terahertz spectroscopic detection.

Abstract

Terahertz time-domain spectroscopy (THz-TDS) based on air-plasma generation and balanced air-biased coherent detection offers gap-free broadband coverage, but individual continuous-scan traces are strongly affected by pulse-to-pulse fluctuations and electronic noise. Reaching a useful signal-to-noise ratio therefore requires averaging multiple traces, which directly increases measurement time. We propose a learned denoising approach that recovers high-quality THz waveforms from as few as one complete continuous delay sweep, referred to here as a single-scan trace. A compact one-dimensional residual U-Net is trained using two complementary strategies: a reference-supervised baseline that maps individual noisy traces to long-average reference waveforms, and a Noise2Noise approach that learns from pairs of independently acquired noisy traces without requiring a clean training target. Averaging the predictions of both models reduces systematic bias and yields a trace-reduction factor of approximately $5.4\times$ at $K=1$, meaning that one denoised trace achieves the reconstruction accuracy of averaging approximately five raw traces. The Noise2Noise model alone achieves $4.9\times$, outperforming both the reference-supervised baseline ($4.6\times$) and classical Wiener filtering ($3.2\times$). These results show that self-supervised learning from repeated noisy measurements can support faster continuous-scan THz-TDS without hardware modification.

Problem

• Terahertz time-domain spectroscopy
• Continuous-scan
• Denoising
• Signal-to-noise ratio
• Measurement time

Innovation

• Self-Supervised Noise2Noise
• THz-TDS Denoising
• Residual U-Net
• Single-Scan Reconstruction
• Trace Reduction