Recursive lattice forms for spectral estimation and adaptive control

Recursive algorithms have been recently developed for modeling time series by pole-zero lattice forms. This development makes it feasible to use lattice forms, which have exhibited excellent numerical behavior, for various applications the areas of adaptive signal processing and adaptive control. A general framework is presented for embedding a large class of signal processing and control problems in a multichannel auto-regressive whitening filter. This framework makes it possible to develop lattice forms for various applications including self-tuning regulators, the adaptive line enhancer, a generalized adaptive noise canceller and spectral estimation. The square-root normalized lattice form and its properties are briefly described. Some of the features of these algorithms are: fast tracking of time varying parameters, simultaneous estimation of models of different orders, computational efficiency and robust numerical behavior.