An application of stochastic automata models to the design of adaptive filters

The design of IIR adaptive filters is considered. The authors present a method composed of two feedback loops an adaptive (LMS) loop and a learning loop that contains a stochastic learning automaton. Preliminary results from simulated examples suggest that this novel approach has a potential for application in many signal processing problems where classical methods may not converge to the global minimum, or give biased or unstable results. The advancement in parallel processing hardware technology such as the availability of high speed-large memory capacity digital signal processors makes the use of learning techniques attractive