Bayesian computational methods for sparse audio and music processing

Simon J. Godsill, Ali Taylan Cemgil, Cédric Févotte, Patrick J. Wolfe

Abstract: In this paper we provide an overview of some recently developed Bayesian models and algorithms for estimation of sparse signals. The models encapsulate the sparseness inherent in audio and musical signals through structured sparsity priors on coefficients in the model. Markov chain Monte Carlo (MCMC) and variational methods are described for inference about the parameters and coefficients of these models, and brief simulation examples are given.