To lower the equalization complexity, the channel matrices are approximated to be banded in both domains.
The eigenvector matrices of these Toeplitz matrices are approximated by submatrices of DFT matrix when the number of antennas is large [19, 20].
Considering the one-ring scattering model, the azimuth and elevation correlations are characterized by Toeplitz matrices, and the eigenvector matrices of these Toeplitz matrices are approximated by submatrices of DFT matrix when the number of antennas is large [19, 20].
However, the rotation matrices are approximated by using a single Gauss Newton step with a fixed updating step length, which can lead to a considerable performance drop in the rotation reconstruction if no proper metric on the manifold is defined.
For this model, North South and East West gradients (as well as their associated kernel matrices) are approximated by simple first differences (no division with the spherical distance), in a similar fashion to how the gradients are commonly treated for modeling, see, for example, Finlay et al.
({mathrm {SM}}_{{Gamma ^{(1)}}delta }) this model is equivalent to ({mathrm {SM}}_{{Gamma ^{(1)}}}) except that the gradients involved in (varvec{Gamma }^{(1)}) (and the associated gradient kernel matrices) are approximated by simple data differences (no division with the spherical distance).
In the LM approach, the Hessian matrix is approximated by.
The Hessian matrix is approximated, using a set of box-type filters.
In this case, the covariance matrix is approximated by C oqam = σ w 2 + θ 2 I NL (105).
In this method, the full elements of CFR matrix is approximated by a banded matrix so as to enable the employment of parallel block inverse matrix algorithm.
