1998•Electronics LettersRequires access
Modelling signals of arbitrary kurtosisfor testing BSS methods
Vicente Zarzoso, Asoke Kumar Nandi
Open publisher page 7 citations
Abstract
It is shown how a very simple distribution, the Bernoulli random variable, exhibits all possible normalised kurtosis values when varying the probability of its two events. This result means that signals of any normalised kurtosis may be modelled by this distributiori, which is of relevance within the framework of blind source separation based on fourth-order cumulants.