Blind decision-feedback equalization of underwater acoustic channels

Blind equalization algorithms for linear FIR filters cannot usually cope with the rapid fading and significant distortion that are frequently observed in underwater acoustic channels. Previous results indicate that performance may be greatly improved by adopting simple nonlinear receivers which have been commonly used in nonblind equalization. In this paper, a predictive decision-feedback structure that uses a super-exponential blind linear equalizer as a forward filter is analyzed. Its rapid convergence to an approximate Wiener solution allows near optimal mean-square performance to be attained with a relatively small number of input samples. The feedback filter is a linear predictor that can be efficiently adapted with the recursive least-squares algorithm. Using these design guidelines, the whole receiver becomes fast and self-tuning, in the sense that it does not require ad-hoc setting of critical convergence parameters. Experimental results using simulated and real data illustrate the effectiveness of this approach.