Design of an optimum Poincaré plane for extracting meaningful samples from EEG signals.

Biosignals are considered as important sources of data for diagnosing and detecting abnormalities, and modeling dynamics in the body. These signals are usually analyzed using features taken from time and frequency domain. In theory' these dynamics can also be analyzed utilizing Poincaré plane that intersects system's trajectory. However' selecting an appropriate Poincaré plane is a crucial part of extracting best Poincaré samples. There is no unique way to choose a Poincaré plane' because it is highly dependent to the system dynamics. In this study, a new algorithm is introduced that automatically selects an optimum Poincaré plane able to transfer maximum information from EEG time series to a set of Poincaré samples. In this algorithm' EEG time series are first embedded; then a parametric Poincaré plane is designed and finally the parameters of the plane are optimized using genetic algorithm. The presented algorithm is tested on EEG signals and the optimum Poincaré plane is obtained with more than 99% data information transferred. Results are compared with some typical method of creating Poinare samples and showed that the transferred information using with this method is higher. The generated samples can be used for feature extraction and further analysis.