Quadratically and linearly constrained statistically dependent time-frequency kernel design

Time-frequency kernel design is formulated as a minimum variance constrained minimization problem. The minimized cost function is quadratic and depends on the second order statistics of the underlying process. In addition to the original constraints leading to the marginal and the support properties, quadratic constraints are employed to add flexibility in kernel design and to allow the user to compromise between the kernel statistical performance under two or more processes. Computer simulations are presented which illustrate statistical performance of the t-f kernel under the original and the additional linear and quadratic t-f constraints.<
>