Effect of equalities in RRI time series on permutation entropy under different emotional states

Heart rate variability (HRV) has been subject to high attention in the field of emotion recognition as it can reflect inherent information objectively and accurately. Various methods of nonlinear dynamics and complexity measures have been utilized for HRV analysis, including fractal dimension, Lyapunov exponents and entropies. Permutation entropy proposed by Bandt is a simple and fast algorithm to investigate complex dynamical systems. It symbolizes the time series and focus on the relative magnitude of data. Therefore, equalities in series should be considered carefully. In this paper, two methods of dealing with equalities have been applied to RRI time series in three emotional states: ranking equal values according to their order of appearance, or eliminate equalities by adding a little random noise. Results show that PE values derived from these two methods are significantly different. The significant differences are demonstrated under all three emotional states.