Blind and Semiblind Channel Estimation - Nonlinear Signal and Image Processing: Theory, Methods and Applications - page 283

8.6.4 Blind Channel Estimation for CDMA

The blind equalization/identification methods for CDMA proposed so far in
the literature employ second-order statistics of the received sampled signals.
The complexity of the HOS-based methods would significantly increase the
already high computational complexity. In the context of long codes, correla-
tion matching, subspace estimation, and maximum likelihood methods have
been used.
In Reference 57, both blind and pilot-aided channel estimation is consid-
ered in the context of long-code systems. A two-dimensional RAKE receiver
is proposed in a scenario where the base station is equipped with multiple
antennas. At the output of each antenna array element, there is a linear FIR
filter for each user. The filter coefficients for the user of interest are adjusted
so that their combined output produces the symbol-modulated spreading se-
quence. The MMSE equalizer coefficients are found using the Wiener-Hopf
equation where the cross-correlation vector is obtained from the correlation
matrices of the pre- and post-despread signal. A principal component algo-
rithm is proposed for blind estimation of the cross-correlation vector. The
blind identification relies on the property that the pre- and post-despreading
covariance matrices of the antenna outputs are different only by a rank-one
matrix defined by the desired cross-correlation vector. A deterministic least-
squares approach to blind equalization is developed, as well, but only for the
case of an underloaded cell.
Related subspace techniques for channel estimation in CDMA with aperi-
odic spreading codes have been proposed. 58 , 59 Apreprocessing stage using
pseudo-inversion of the code matrix is employed in Reference 58. Channel is
identified using a noise subspace method. Both blind and data-aided methods
are presented and their performance is compared with the theoretical per-
formance bounds. Multicode transmission and multiantenna receivers using
subspace channel estimation are considered in Reference 59. A partial de-
spreading preprocessing stage leads to significant savings in computational
complexity and improved robustness to code selection in comparison to that
described in Reference 58.

8.6.5 Blind MUD
With the traditional RAKE receiver, 60 the user of interest is coherently solving
the multipaths of the received signal. It uses the maximum ratio combining
(MRC) to deliver optimal performance. Unfortunately, the RAKE receiver is
very sensitive to multiuser interference (MUI). Therefore, more powerful tech-
niques are needed to cope with the MUI. The MUD techniques 27 exploit both
the code knowledge and the channel characteristics of each user to improve
the overall performance.
Madhow 61 has presented an excellent review of the blind adaptive tech-
niques for channel identification and interference suppression in CDMA.