Blind and Semiblind Channel Estimation - Nonlinear Signal and Image Processing: Theory, Methods and Applications - page 284

He classifies the receivers based on the amount of prior knowledge avail-
able. Adaptive interference suppression techniques are also considered by
Honig 63 and Poor and Wornell. 62 An overview of the multiuser receivers for
short code CDMA may be found in Reference 63, where different approaches
to blind and data-aided methods are compared.
Honig proposed the first blind MUD technique for short-code CDMA; see
Reference 63. The constrained minimum output energy approach related to
MVDR beamforming was developed. The desired user's channel response
was assumed known. Madhow 61 combined blind MUD with blind channel

estimation.
A subspace method for blind MUD and channel identification has been
proposed by Wang and Poor. 64 When certain conditions are fulfilled, such
as the channel matrix has a full column rank, the channel matrix is shown
to be blindly identifiable up to a nonsingular ambiguity matrix. A multiuser
extension of the well-known algorithm of Reference 35 is developed, and
using the knowledge of the user spreading codes, the ambiguity inherent
to blind identification may be solved in both uplink and downlink cases.
In uplink, the users' propagation delays are assumed known. In the special
case of flat fading, a method for estimating the propagation delays is derived
as well. Projecting the users' estimated channels into the signal subspace,
MMSE and zero-forcing detectors are also derived and their robustness is
analyzed. Torlak and Xu 65 have proposed a blind subspace method for blind
channel identification in asynchronous CDMA. They show that the subspaces
of the data matrix contain sufficient information for unique determination of
channel. There are typically three steps in the subspace MUD computations:
channel delay estimation, the delayed components' gain estimation, and the
actual MUD projection. All steps exploit the signal subspace of the received
sequence.
A blind MUD in dispersive channels has been proposed. 66 , 67 Only the de-
sired user's code needs to be known. The receiver is found as a solution to a
special MMSE cost function minimization in two stages. The signal subspace
as well as the code subspace properties are used to define the cost function,
thus allowing for improved performance at low SNR. Several receive anten-
nas may be used.
Papadias and Huang 68 have proposed two linear adaptive space-time ap-
proaches to MUD in CDMA. The optimum space-time MMSE MUD is first
derived and its bit error rate (BER) performance is analyzed. Two adaptive im-
plementations of the MMSE detector are developed: a blind multiuser CMA
(MU-CMA) and a trained MU-LMS. The global convergence of the blind
method is proved as well.
If there are multiple antennas available at the transmitter, space-time codes
can be used to improve the system performance. In Reference 69 the properties
of a space-time block coding scheme are used with the spreading code prop-
erties for canceling various interferences in a CDMA downlink with multiple
antennas at the base station. The increased interuser interference due to the