Misconvergence and stabilization of adaptive IIR lattice filters

We take a new look to adaptive IIR lattice filters in order to devise algorithms that preserve the stability of the corresponding direct form schemes. By analyzing the local properties of stationary points, a transformation achieving this goal is suggested, which gives algorithms that can be efficiently implemented. Application to the Steiglitz-McBride and SHARF algorithms is presented. By contrast, it is shown that previous lattice versions of these algorithms may fail to preserve stability of stationary points.