Analysis of Chaotic Signals: Non-linear Methods versus Neural Networks

Applications of Non-linear Methods and Neural Networks in the analysis of chaotic signals are compared in the paper. Results of time series analysis by non-linear methods are illustrated by computations of Lyapunov exponents and correlation dimension. Abilities of Neural networks are demonstrated in reconstruction of chaotic attractors, in generation of chaos and in the classification and modelling of a selected chaotic signal. A practical deployment of developed methods and programs has been included in diagnosis of cardiovascular system using signals of Electrocardiograms (ECG) and of Heart Rate Variability (HRV). For the data support there have been used databases MIT-BIH, Fantasia and Physio Bank Database.