Blind and Semiblind Channel Estimation - Nonlinear Signal and Image Processing: Theory, Methods and Applications - page 287

introduced in the literature rely on pilot sequences. Blind methods have also
received attention in OFDM because they save bandwidth, reduce overhead
signaling, and allow tracking of slow-channel variations. A number of blind-
channel estimation techniques have been developed for OFDM, some of them
based on cyclostationarity 72 or on subspace decomposition that exploits the
CP and zero padding (ZP) structure of OFDM. 73 One problem that may arise
is related to HOS-based blind methods. In OFDM, due to the high number
of subcarriers, the transmit signals are close to Gaussian. Hence, application
of blind separation types of algorithms, which rely on HOS, is difficult. Blind
techniques based on SOS have been proposed instead. 73 , 74
The algorithm presented in Reference 73 is based on subspace decomposi-
tion and takes advantage of the inherent redundancy introduced by the CP
in estimating the channel in a blind manner. This technique is insensitive to
the constellation size, unlike the decision-directed techniques, which suffer
from performance degradation when a higher-order constellation is used. It
also guarantees the channel identifiability regardless of the channel zero's
location when the entire noise subspace is considered. A semiblind version
was introduced, 73 which improves convergence speed of the algorithm. The
quality of the initial channel estimates at the beginning of the frame is also
improved.
The method proposed in Reference 74 solves the problem of channel es-
timation and equalization in a blind manner for a MIMO system. SOS are
used to identify the channel on a subchannel-by-subchannel basis. This tech-
nique does not have restrictions to channel zeros and exhibits low sensitiv-
ity to stationary noise. An upper bound on the channel order is assumed
known.
In OFDM transmission, CP allows for a simple one-tap equalization scheme.
The data rate can be increased by reducing the CP length. This may cause se-
vere ISI, and more powerful equalization schemes are needed to mitigate this
effect. Another way of reducing ISI is obtained by using ZP instead of CP. This
can be considered to be a form of periodic precoding. A blind channel esti-
mator exploiting the CP has been proposed. 75 The method uses the algebraic
structure of the signal model, which is a function of the channel.
Based on the observation that the multicarrier system turns a single wide-
band frequency-selective channel into a set of correlated narrowband flat-
fading channels, Luise et al. 76 propose a blind channel estimation-detection
scheme that exploits the frequency correlation between neighboring subcar-
riers. Channel equalization is accomplished by means of a trellis decoder on
a block-by-block basis and is shown to perform significantly better than a
differential encoding-decoding scheme.
Compared to the single-transmit antenna case, the channel estimation
for multiple-transmit antenna systems is made more difficult by the fact that
the received signal is a superposition of the transmitted signals from diff-
erent antennas. The problem of channel estimation was investigated in
SIMO 77 and MIMO 74 scenarios. The SIMO type of scenario can be obtained by