A high-performance training-free pipeline for robust random telegraph signal characterization via adaptive...

Area of Science:

• Physics
• Chemistry
• Biology
• Materials Science

Background:

• Random telegraph signals (RTS) reveal temporal fluctuations in various systems.
• Analyzing RTS is complex due to noise and multi-level signals.
• Accurate RTS characterization is crucial for understanding microscopic processes like charge trapping.

Purpose of the Study:

• To develop a high-throughput, training-free signal processing pipeline for RTS analysis.
• To improve the accuracy and speed of RTS characterization, especially in noisy conditions.
• To provide a scalable and reproducible foundation for autonomous RTS studies.

Main Methods:

• Adaptive dual-tree complex wavelet transform (DTCWT) for denoising with automatic parameter selection.
• Lightweight Bayesian digitization for probabilistic latent-state inference.
• Temporal regularization without iterative optimization for robust state resolution.

Main Results:

• Significantly improved RTS reconstruction accuracy and trap-state resolution.
• Enhanced dwell-time estimation across diverse noise regimes and multi-trap scenarios.
• Achieved up to 83x speedup compared to classical and neural network baselines.

Conclusions:

• The proposed pipeline offers a scalable and reproducible framework for autonomous RTS analysis.
• Demonstrated practical usability and flexibility for real-time and large-scale experimental data analysis.
• Provides a foundation for future extensions to more complex RTS characterization.