Analysis of non-stationary signals by recurrence dissimilarity

Abstract

We propose a new method for testing non-stationary and intermittent signals. Using a basic recurrence plot quantifier, i.e. a recurrence rate, we create a relative measure which is sensitive to changes in the nature of signal. Given the specificity of this measure, we call it recurrence dissimilarity (RD). First, we test it using well-known non-linear systems for which we generate signals with different intermittent characteristics. In addition, the generated signals are disturbed by noise. The effectiveness of our measure is verified by applying different variables and noise levels. The results allow us to draw a number of conclusions concerning the proposed method. Finally, we give examples of using this method for experimental data analysis. We report the results of detecting changes in flowing patterns in two-phase flows and of switching heart modes in the signals recorded in ECG Holter tests.