The weighted least-squares (LS) and total-least-squares (TLS) solution
s to a perturbed system of linear equations are shown to be asymptotic
ally equivalent (as the perturbation level goes to zero). An immediate
consequence of this fact is that an optimally weighted LS-ESPRIT meth
od, recently proposed for bearing estimation, achieves the lowest poss
ible asymptotic estimation error variance in the classes of weighted T
LS-ESPRIT estimates as well, and hence it is statistically efficient f
or Gaussian perturbations.