132)
where NI is the dimension of the interference vector space of ?†R
IN.
From this spectral decomposition many insightful array weights can be derived and used under
different conditions. Below we will list three of them:
1. MMSE signal subspace. Here we use only the eigen vectors that correspond to the signal subspace,
which is of size Ns. The reasoning behind this weight is as follows. The cross-correlation vector is
a member of the signal subspace and as such is collectively orthogonal to the interference noise
subspace. Hence the second summation would result in contributing ideally a value of zero. In
practice, it is not zero, but rather a small value and so by forcing the property performance
improvements can be observed.
(5.133)
2. MSINR noise subspace. This technique is also referred to as the Eigen Cancellor [71].
The reasoning behind this array weight is as follows. Since we are dealing with the RIN covariance
matrix the noise subspace is collectively orthogonal to the interference subspace. Hence in
using this noise vector space we can attempt to cancel/suppress the interference that belongs to the
interference subspace. That is why the second half of the summation is used.
(5.134)
3.
Pages:
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456