Precise modeling of reverberant room responses using wavelet decomposition and orthonormal basis functions

Precise modeling of a room impulse response (RIR) using a small number of parameters is required in many audio processing applications. However, this is challenging since room responses may be tens of thousands of taps long in practice and vary greatly as the source and microphone locations are slightly changed. In this paper a subband multichannel method for accurate modeling of long RIRs is proposed, which is computationally efficient and robust against RIR variations. A dual-tree complex wavelet packet transform is utilized to decompose a multichannel RIR into aliasing-free subband signals, and low order adaptive Kautz filters are designed to model subband signals using the acoustical poles common to the RIR channels (common acoustical poles (CPs)). A least-squares algorithm is introduced to efficiently estimate the CPs at each subband. The algorithm precisely estimates the CPs after a low number of iterations and unconditionally guarantees the stability of the estimated poles. Experimental results demonstrate that the proposed method accurately models the room responses while exhibiting robustness against room response variations caused by changing the source and microphone locations. © 2018 Audio Engineering Society. All rights reserved.