Nonlinear Blind Deconvolution Based on Kurtosis Criterion and Decision Directed Algorithm

A new blind deconvolution algorithm for Weiner model is proposed, based on kurtosis criterion and decision directed. Through analyzing when maximum kurtosis is used to resolve nonlinear blind deconvolution problem, it is found there exist some disadvantages, such as too many local optimum values and large residual error. So the decision directed least mean error is introduced in the cost function, and the number of local optimum values can be reduced and residual error is decreased. To overcome the drawback of traditional gradient search approaches, likely falling into local minimum, the real coded genetic algorithm is adopted to search the optimum solution. Simulation results demonstrate this algorithm not only has fast convergence rate and high accuracy, but also can greatly improve the output signal noise ratio.