Fractionally spaced blind equalization: CMA versus second order based methods

It is now well established that most of the blind fractionally spaced channel estimates based on the exclusive use of the second order statistics of the observation have poor performance when the excess bandwidth of the received signal is reduced. Previous papers proposed to use covariance matching approaches, and showed that a significant improvement of the mean square error of the channel estimate is possible. The purpose of this paper is to evaluate more precisely the potential of these methods. For this, we propose to compare the performance of a Wiener equalizer based on the optimally weighted covariance matching estimate of the channel (known as the best second order statistics based estimate) with the most standard higher order statistics based method, i.e. the (block) fractionally spaced CMA equalizer. It is shown that the CMA outperforms quite significantly the second order statistics based Wiener equalizer.