Joint structured channel and data estimation over time-varying channels

This paper describes an adaptive maximum likelihood sequence estimation (MLSE) receiver based on a structured linear channel model. Specifically, known prior information about the transmit filter is used to obtain a structured channel model linearly parameterized by a time-varying vector. We show that the total oversampled FIR channel vector lies within the subspace of a matrix associated only with the samples of the transmit pulse shape. The resulting structured channel model has typically less numbers of unknown parameters than the conventional unknown FIR channel model. Joint structured channel and data estimation (JSCDE) is done based on a per-survivor processing (PSP) approach. A reduced complexity scheme for JSCDE is also proposed that uses delayed decision feedback for each survivor in a reduced trellis.