M. C. Ruiz Abellón, A. Guillamón Frutos, J. S. Cánovas Peña
An alternative procedure to detect structural changes in time series is proposed, which is based on the number of different permutations that appear in a data series. The efficacy of the method is shown through simulated data and the influence of the embedding dimension (permutations length) is also pointed out. Some applications to real data are included, such as detecting epileptic seizures from EEG signals or the awake or sleeping state in depth of anesthesia.
Palabras clave: time series, permutations, structural changes
Programado
XC1a Pósters (Estadística)
18 de abril de 2012 12:00
Salón Madrid