Sparse Representation of Sound Field Based on K-SVD

Abstract

Sparse representation has been applied to nearfield acoustic holography in order to provide an accurate reconstruction with a small number of sampling points. However, the performance of the application is based on the sparsity of the signal to be reconstructed. In this study, K-SVD dictionary learning method is used to construct a basis of the sound field and the sound field can be sparsely represented by the learned dictionary. The samples of K-SVD are the sound fields obtained by simulations rather than those measured in practice, by taking the advantage of the equivalent source method. Compared with other sparse bases used in the existing nearfield acoustic holography, the dictionary learned by K-SVD is more flexible and the prior of the sound field is not necessary. The numerical simulation results demonstrate the validity of advantage of the proposed method.