Employing adaptive functions and maximum entropy principle for nonlinear blind source deconvolution