Autoencoder-Based Parameter Estimation for Superposed Multi-Component Damped Sinusoidal Signals

TL;DR

This paper introduces an autoencoder-based approach for accurately estimating parameters of multi-component damped sinusoidal signals in noisy conditions, outperforming traditional methods.

Contribution

The study presents a novel autoencoder method that effectively estimates multiple signal parameters even with rapid decay, noise, and superposition, and explores training data effects.

Findings

High accuracy in parameter estimation for challenging signals

Robustness to less informative training data

Effective waveform reconstruction in noisy environments

Abstract

Damped sinusoidal oscillations are widely observed in many physical systems, and their analysis provides access to underlying physical properties. However, parameter estimation becomes difficult when the signal decays rapidly, multiple components are superposed, and observational noise is present. In this study, we develop an autoencoder-based method that uses the latent space to estimate the frequency, phase, decay time, and amplitude of each component in noisy multi-component damped sinusoidal signals. We investigate multi-component cases under Gaussian-distribution training and further examine the effect of the training-data distribution through comparisons between Gaussian and uniform training. The performance is evaluated through waveform reconstruction and parameter-estimation accuracy. We find that the proposed method can estimate the parameters with high accuracy even in challenging setups, such as those involving a subdominant component or nearly opposite-phase components, while remaining reasonably robust when the training distribution is less informative. This demonstrates its potential as a tool for analyzing short-duration, noisy signals.