Audio Signal Processing and Coding

8.7 SUBBAND CODING WITH IIR FILTER BANKS

Although the majority of subband and wavelet audio coding algorithms found in the literature employ banks of perfect reconstruction FIR filters, this does not preclude the possibility of using infinite impulse response (IIR) filter banks for the same purpose. Compared to FIR filters, IIR filters are able to achieve similar magnitude response characteristics with reduced filter orders, and hence with reduced complexity. In the multiband case, IIR filter banks also offer complexity advantages over FIR filter banks. Enhanced performance, however, comes at the expense of an increased sensitivity and implementation cost for IIR filter banks. Creusere and Mitra constructed a template subband audio coding system modeled after [Lokh92] to compare performance and to study the tradeoffs involved when choosing between FIR and IIR filter banks for the audio coding application [Creu96]. In the study, two IIR and two FIR coding schemes were constructed from the template using a structured all-pass filter bank, a parallel all-pass filter bank, a tree-structured QMF bank, and a PQMF bank. Beyond this study, IIR filter banks have not been widely used for audio coding. The application of IIR filter banks to subband audio coding remains a subject that is largely unexplored.