Finite impulse response (FIR) adaptive filters are now widely used for equalization of communication channels. However in many applicatios, such as in multipath environments a more general infinite impulse response (IIR) filter is required to achieve optimal performance. These are much more difficult to design than their FIR counterparts, as there is, for instance an added requirement that the resulting filter must be stable. In addition, the adaptation of an IIR filter will, in general, require optimization of a cost function with local minima. In this paper we have shown how these difficulties can be overcome by using our proposed approach to adaptive IIR filtering. We employ genetic algorithms (GAs) on a parallel form complex adaptive IIR structure to evolve a population of filter objects that adapts over a period of time. Unlike existing methods, our method provides some interesting features including guaranteed filter stability, the ability to equalize complex channels, searching for a global optimum is the error surface having multiple local minima, and reduced complexity.
展开▼