Location and Spectral Estimation of Weak Wave Packets on Noise Background

TL;DR

This paper presents a novel method for locating and estimating the spectrum of weak wave packets within noisy backgrounds, utilizing an optimized autoregressive model and new model order determination techniques.

Contribution

It introduces a new approach for model order selection and noise dispersion estimation, enhancing detection of nonstationary wave packets at very low SNR levels.

Findings

Effective detection of wave packets at SNR as low as -20 dB.

Accurate estimation of noise dispersion close to real values.

Use of dynamic invariant analysis improves localization accuracy.

Abstract

The method of location and spectral estimation of weak signals on a noise background is being considered. The method is based on the optimized on order and noise dispersion autoregressive model of a sought signal. A new approach of model order determination is being offered. Available estimation of the noise dispersion is close to the real one. The optimized model allows to define function of empirical data spectral and dynamic features changes. The analysis of the signal as dynamic invariant in respect of the linear shift transformation yields the function of model consistency. Use of these both functions enables to detect short-time and nonstationary wave packets at signal to noise ratio as from -20 dB and above.