Assessing the similarity between time series using a Wavelet transform: Application and interpretability aspects

Abstract

This work presents a simple and interpretable measure to evaluate the similarity between biosignal time series. Combining the Haar wavelet with the Karhunen Loève transforms, the proposed similarity measure is particularly appropriated to deal with noisy signals, with signals that are not time aligned as well as to recognize similar trends. When applied to an indexing scheme, an iterative formulation enables a very efficient computational implementation. Experimental and simulation results using blood pressure signals collected by a telemonitoring platform (TEN-HMS) show the effectiveness of the proposed scheme.