The block underdetermined covariance (BUC) fast transversal filter (FTF) algorithm for adaptive filtering

Fast algorithms in transversal filter form (FTF) developed for an adaptive filtering strategy consisting of treating consecutive blocks of L data are discussed. The resulting BUC FTF algorithm can be situated in between the normalized least-mean-square (NLMS) algorithm for L=1 and one extreme case of the block-processing FTF algorithm for L=N. A projection mechanism onto a subspace of dimension L renders their convergence less sensitive to the coloring of the input signal spectrum than is the case for the NLMS algorithm. Their underdetermined LS character endows them with relatively fast tracking characteristics. Relations to existing algorithms and various implementations with varying degrees of complexity are also discussed