Iterative blind linear equalization in time-varying dispersive channels

We present a novel approach to blind linear equalization, and data detection, for efficient, uncoded transmission over a frequency selective Rayleigh fading channel. At each symbol interval, a primary data estimate is made based on the constrained linear minimum mean square error (MMSE) criterion, and this is used by the subsequent channel estimation. As a result, unlike the known Kalman-based tracking algorithm, there is no delay between the channel estimation and the channel estimates needed for the current equalizer adaptation. The performance is evaluated by simulation, allowing fair comparison with: the benchmark of equalized, coherent detection; a known blind system based on Kalman filtering which employs channel tracking with delay; and for the flat fading case, an optimized pilot symbol assisted modulation (PSAM) system.