Characterization of dynamical systems under noise using recurrence networks: Application to simulated and EEG data

• We study the influence of noise on the recurrence network measures C and L.
• We propose the application of C and L to healthy and epileptic EEG data.
• The influence of noise can be minimized by increasing the recurrence rate.
• Measures C and L can describe the structural complexity of EEG data.
• In case of epileptic EEG, C performs better than L.

In this letter, we study the influence of observational noise on recurrence network (RN) measures, the global clustering coefficient (C) and average path length (L  ) using the Rössler system and propose the application of RN measures to analyze the structural properties of electroencephalographic (EEG) data. We find that for an appropriate recurrence rate (RR>0.02RR>0.02) the influence of noise on C can be minimized while L is independent of RR for increasing levels of noise. Indications of structural complexity were found for healthy EEG, but to a lesser extent than epileptic EEG. Furthermore, C performed better than L in case of epileptic EEG. Our results show that RN measures can provide insights into the structural properties of EEG in normal and pathological states.