Efficient Identification of Acoustic Linear Systems

The identification of acoustic linear systems is a critical issue in numerous applications related to acoustic en-vironments. A major difficulty that arises in this context is the long length of impulse responses. In this paper, we present an efficient method to address this issue, using the nearest Kronecker product decomposition of the impulse response, along with low-rank approximations, which can be further improved by using a proper permutation matrix. As a result, we develop an iterative Wiener filter using this method, with superior performances with respect to the conventional Wiener filter, especially in the case with small amount of data available for the estimation of the required statistics. Simulations performed in the context of stereophonic acoustic echo cancellation support the advantages of the proposed solution.