Loudspeaker impulse responses require a large number of data points (e.g., 8192). For many applications it is beneficial to reduce this order generating fully parametric response models. This can be achieved by fitting pole-zero (ARMA) models by Prony analysis, a method suffering from poor definition in the low-frequency region. Here, this problem is solved by signal bandsplitting, achieving perfect response modeling with 256 : coefficients.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.