Numerical spread: Past results and future directions

In previous work the numerical spread was introduced as a means of quantifying the stationarity of a signal. In this paper we will review those results and propose future research that addresses two of the major shortcomings of the method: robustness and computability. It is proposed to apply the adaptive local cosine covariance estimation techniques developed by Mallat et al. to compute the covariance. Given this estimate and the segmentation inherited from the local cosine basis we will then compute a spread on these regions. Thus providing a method to improve robustness and address the problem of computability.