Efficient adaptive filtering in subbands using IIR filter banks

This paper deals with a structure for implementing adaptive filters using orthogonal, lossless filter banks. Previously, only FIR filter banks have been used for this structure. We demonstrate that also computationally efficient IIR filter banks can be used even if the adaptive filter to be implemented is an FIR filter. We show that the computational advantages obtained using IIR filters do not lead to performance degradations. It is also demonstrated that the convergence properties of the subband adaptive system-in contrast to the least mean square (LMS) algorithm and transform domain adaptive algorithms based on for example the discrete cosine transform (DCT), is robust with respect to the properties of the input signal to the adaptive filter.