Time-varying channel equalization using multiresolution analysis

The equalization problem for rapidly varying channels is addressed, using multiresolution analysis tools. The time variations of the channel are modeled directly by expanding each time-varying (TV) tap coefficient onto a P-band wavelet basis. Only the low-resolution signal is then retained as a natural approximation of the channel's trajectory. The choice of the wavelet basis so that the approximation is optimal in the mean square sense is discussed. An algorithm based on higher-order statistics, which estimates the wavelet coefficients from output-only data, is proposed. Once the channel has been estimated, the Viterbi algorithm is used to complete this blind equalization approach.<
>