Comparación de métodos de análisis de señales en el dominio del tiempo-frecuencia, considerando la resolución en tiempo, resolución en frecuencia, detección y carga computacional

Abstract

The most relevant results of the comparison of four signal analysis methods in the time-frequency domain are presented: Short Time Fourier transform, Wigner-Ville distribution, Wavelets and Matching Pursuit, using artificial signals, in order to find the advantages and disadvantages of each of these methods in terms of frequency resolution, time resolution, detection and computational load.

Three experiments were performed. The signals were constructed in such a way that they had increasingly stringent conditions of frequency resolution (less and less frequency separation between components) and detection (decrease in the amplitude of the components); for the resolution in time different start and end times were considered for each component. The results show that the best method in terms of frequency resolution, detection and computational load is the Short Time Fourier Transform. On the other hand, using Wavelets the best resolution in time is obtained, in addition with the Wavelet Morlet a diagram called “Fine Scale Coefficients” is obtained that allows a more precise resolution analysis in time.

Additionally, a new method of signal analysis has been proposed taking advantage of the previously mentioned methods and combining their best features. It was possible to obtain very good results of resolution in time, resolution in frequency, detection and computational load.