On adaptive interpolated FIR filters

The computational complexity and memory load of the adaptive finite impulse response (AFIR) filter are significant for a large filter size. The adaptive interpolated FIR (AIFIR) filter, which uses a sparse adaptive filter followed by an interpolator, has been shown to be a better alternative. However, when the AIFIR filter is implemented, the coefficients of the interpolator must be designed in advance based on prior information about the application at hand. Such information is not always available and the design of a proper interpolator is sometimes difficult. In this paper we introduce a new structure called double adaptive interpolated FIR (DAIFIR) filter in which the fixed interpolator is replaced by an adaptive filter of the same length. We show by means of simulations that the behavior of the proposed structure is close to the behavior of the AIFIR having a proper designed interpolator. In situations in which a fixed interpolator cannot be designed in advance the DAIFIR might be a good alternative.