Filter tuning algorithm with compressed reflection characteristic by Daubechies D4 wavelet transform

This work presents how to improve an algorithm based on artificial neural network (ANN) for microwave filter tuning. Sets of ANN learning vectors which contain scattering filter characteristics and corresponding tuning elements deviations are used in the concept. The main idea is to transform filter characteristics to wavelet Daubechies D4 representation before ANN training. Experimental results have shown that usage of truncated D4 filter characteristics allows us to minimize ANN topology and thus has significant influence on ANN training time, whilst ANN generalization ability remains at a similar level.