Improving Robustness of Spectrogram Classifiers with Neural Stochastic Differential Equations

TL;DR

This paper proposes a novel approach using Neural Stochastic Differential Equations to enhance the robustness of spectrogram classifiers against noise and perturbations, aiming to improve signal classification in noisy, real-world environments.

Contribution

It introduces a neural stochastic differential equation framework specifically designed to improve robustness of spectrogram-based classifiers in noisy signal processing tasks.

Findings

Enhanced robustness of classifiers against noise

Improved accuracy in low signal-to-noise ratio scenarios

Potential applications in critical infrastructure monitoring

Abstract

Signal analysis and classification is fraught with high levels of noise and perturbation. Computer-vision-based deep learning models applied to spectrograms have proven useful in the field of signal classification and detection; however, these methods aren't designed to handle the low signal-to-noise ratios inherent within non-vision signal processing tasks. While they are powerful, they are currently not the method of choice in the inherently noisy and dynamic critical infrastructure domain, such as smart-grid sensing, anomaly detection, and non-intrusive load monitoring.