Features based on instantaneous frequency for seismic signals clustering

Abstract

Seismic signals discrimination is a multidimensional problem since recorded events may vary in terms of type, location, energy, etc. Recently, two discrimination features based on instantaneous frequency (IF) were proposed by the Authors. The first of these features is determined by distribution of the signals’ first Intrinsic Mode Function’s (IMF) IF. The second one is a particular simplification of the previous one as it gives information about the most frequently occurring instantaneous frequency in the considered first IMF. In order to exhibit features’ potential in distinguishing of seismic vibration signals, one has to use clustering algorithms. The features were already subjected to k-means algorithm. In this paper we show results of agglomerative hierarchical clustering (AHCA) and compare it with outcomes of k-means. In order to test optimal number of clusters, method based on average silhouette was accomplished. The results are illustrated by analysis of real seismic vibration signals from an underground copper ore mine.