MSINR weighted subspace. Here we use the entire vector space of the covariance matrix, but they
are weighted by an eigenvalue modified by the function, f(x) [72]. The reasoning behind this array
weight is as follows. Reduced rank techniques tend to degrade faster as the number of interferers
increases or the dimension is incorrectly estimated. We wish to retain the entire vector space but
weigh then appropriately. By this we mean, each eigenvalue will be inspected and we will determine
whether it is associated to a vector in the interference subspace or noise subspace. If they
belong to the noise subspace, we will force their respective eigenvalues to equal, .
(5.135)
Interference Study. In this section, we will present simulation results of some of the abovementioned
array weights. The received signal is represented as
(5.136)
where _ his an M 1 desired signal channel vector, hIi
is the ith interfering signal channel vectors, and
d is the desired signal, di is the ith interference signal, and n is an M 1 noise vector. P is used as the
number of equal power CCI interferers in the channel. We can define the SINR as follows with INR
interference to noise ratio.
(5.137) SINR
SNR
1 aP
j1
INR
x h # d aP
i1
hIi
# di n
w^
MSINRWSS caM
i1
1
f (li) # li
# vi # v*i
drxd
s2n
w^
MSINRNS c aM
iNI1
1
s2n
vi # v*i d # rxd
(w^
MSINREC)
w^
MMSESS caN
s
i1
1
li
# vi # v*i
d # rxd
w^
MSINR caN
I
i1
1
li
# vi # v*i
aM
iNI1
1
li
# vi # v*i
d # r^
xd
278 CHAPTER FIVE
Assuming the desired signal, interfering signals and noise are all independent, we can write the
expression for the signal plus interference plus noise covariance matrix.
Pages:
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457