Low-Complexity Iterative Estimators of Quasi Static Frequency Selective Channels

This paper proposes iterative estimators of quasi static frequency selective channels. Their properties (in terms of expectation and mean-squared estimation error) as well as complexity are discussed. Their average performance is also analytically approached. For the sake of low complexity, the derived estimators are either linear or affine in the vector of observations at the channel output. Such estimators are referred to as iterative because they use soft information on data symbols fed back by the soft-in/soft-out (SISO) channel decoder of a turbo equalizer. Because involved in this iterative process, they perform better at low signal-to-noise ratios (SNRs) than standard (i.e. non iterative) channel estimators using the same number of pilot symbols. Conversely, iterative estimators require a smaller number of pilot symbols to achieve the same performance as standard estimators.