Improving Cramér-Rao Bound With Multivariate Parameters: An Extrinsic Geometry Perspective

We derive a vector generalization of the square root embedding-based curvature-corrected Cram\'er--Rao bound (CRB) previously considered by the same author in \cite{srk} with scalar parameters. A \emph{directional} curvature correction is established first, and sufficient conditions for a conservative matrix-level CRB refinement are formulated using a simple semidefinite program. The directional correction theorem is rigorously illustrated with a Gaussian example.