Error Weighting and Frequency Warping

Audio filter designs typically benefit from an error weighting function that weights frequencies according to their audibility. An oversimplified but useful weighting function is simply $ 1/\omega$, in which low frequencies are deemed generally more important than high frequencies. Audio filter designs also typically improve when using a frequency warping, such as described in [88,78] (and similar to that in §I.3.2). In principle, the effect of a frequency-warping can be achieved using a weighting function, but in practice, the numerical performance of a frequency warping is often much better.


Next Section:
Stability of Equation Error Designs
Previous Section:
Equation Error Formulation