Eigenanalysis-based blind methods for identification, equalization, and inversion of linear time-invariant channels

Some blind algorithms are presented for estimation of equalizers for linear time invariant (LTI) channels from the eigenvectors of certain rank-one matrices constructed from the second-order statistics of the oversampled received signal. It is shown that the channel can also be identified from the same matrices. It can be shown that in multipath dominated environments, equalizers with symbol spread of only one (referred to as inverters) can be used when sufficient diversity is available. Because of the manner in which structure in the channel distortion is exploited, the proposed identification and equalization algorithms are also applicable to this case. For the same reason, the proposed algorithms do not require estimation of the channel memory (only an upper bound is required). Equalizers of desired delay are estimated directly independent of others.