Spectral density estimation of telemetric data by means of a wavelet-based approach

Abstract

The present paper is concerned with a new technique intended for the spectral density estimation of telemetric signals with the help of the wavelet transform. We briefly revise basic information on classical spectral estimates based on the calculation of non-modified and modified Fourier periodogram and estimates obtained via parametric modeling (autoregressive, etc.). The main problems that arise when applying the estimates mentioned are given and the ways of their solution are offered. We introduce a new approach with the use of the wavelet transform and wavelet-thresholding of detailed coefficients in the decomposition. The new approach was tried on various telemetric data reflecting the functioning process of complex dynamic objects (slowly and fast changing signals). As a result the advantages of a newly developed technique become vivid and obvious: greater accuracy and smoothness of a final spectral estimate.