Multitaper reduced interference distribution

Time-frequency distributions (TFDs) belonging to Cohen´s class are usually designed to satisfy the frequency marginal. The conventional frequency marginal is equivalent to the classical periodogram. It is known that the periodogram is not a good spectral estimator. For this reason, Thomson (Proc. IEEE, vol.70, no.9, pp.1055-96, Sept. 1982) introduced a multitaper spectral estimator which treats both the bias and the variance problems inherent to nonparametric spectral estimators. In this paper, we are introducing a new kernel design method which achieves Thomson´s spectral estimate as the frequency marginal. This new method results in a signal-dependent kernel design. The resulting kernel belongs to the class of reduced interference distribution (RID) kernels and therefore this new time-frequency distribution will be called multitaper reduced interference distribution (MT-RID). The performance of this method is compared to the previously introduced multiwindow time-frequency distribution (MW-TFD) (Frazer and Boashash, 1994) through simulations