Multiwindow estimators of correlation

Many algorithms for signal and array processing have embedded within them sample estimates of correlation. In this paper, we prove that the most general symmetric, quadratic, nonnegative-definite, modulation-invariant estimator of correlation is a multiwindow estimator. We establish that multiwindow estimators have the potential to reduce estimator mean-squared error by reducing variance at the expense of controllable bias. When multiwindow estimators are used to solve signal and array processing problems, they have the potential to improve and generalize many standard results