Regularity Changes Detection Using Maximum Approximate Entropy and Particle Swarm Optimization

JUAN FELIPE RESTREPO; GASTÓN SCHLOTTHAUER; MARÍA EUGENIA TORRES

San Carlos de Bariloche

Congreso; XV Reunión de Trabajo en Procesamiento de la Información y Control RPIC 2013; 2013

Universidad Nacional de Rio Negro - Comisión permanente RPIC

Approximate entropy (ApEn) has proved to be a useful tool to discern between different dynamics when there is available short-length noisy data. However, in its estimations, the incorrect parameterselection (embedding dimension m and tolerance r) can undermine its discrimination capacity.Among the family of ApEn statistics, ApEnmax has been proposed as a reliable complexity estimator despite its calculation requires high computational efforts. Here we propose a method that not only improves the discrimination capacity between different dynamics but also can reduce the time spent in calculations. This method, based on the Particle Swarm Optimization algorithm, uses ApEnmax along with the r value at which ApEnmax is achieved (rmax) as discriminating features. We test its performance through simulations with synthetic signals from nonlinear models and with a real electroencephalography signal (EEG). The results show that as well as ApEnmax, rmax can also be used as a feature to discern between dynamics. Further, using together the information provided by both estimators, the method achieves better discrimination capacity than using each one individually. The computational cost is reduced by the inclusion of the Particle Swarm Algorithm.