On the Performance of Time Frequency Distributions in A-Scan Signals Classification

Abstract

In this paper we discuss the performance of different time frequency distributions in characterizing AScan signals in NDT applications. We then introduce a new set of time frequency features that we extract from such distributions. In particular, we propose to extract four signals representing energy and frequency parameters of Ascan signals. We show that the means and spreads extracted from such signals can be used as robust features for classification of A-scans signals. We also show the best discrimination among different classes of A-scans can be obtained using the Gabor transform.