Title :
A Max-Product EM Algorithm for Reconstructing Markov-Tree Sparse Signals From Compressive Samples
Author :
Zhao Song ; Dogandzic, A.
Author_Institution :
Dept. of Electr. & Comput. Eng., Iowa State Univ., Ames, IA, USA
Abstract :
We propose a Bayesian expectation-maximization (EM) algorithm for reconstructing Markov-tree sparse signals via belief propagation. The measurements follow an underdetermined linear model where the regression-coefficient vector is the sum of an unknown approximately sparse signal and a zero-mean white Gaussian noise with an unknown variance. The signal is composed of large- and small-magnitude components identified by binary state variables whose probabilistic dependence structure is described by a Markov tree. Gaussian priors are assigned to the signal coefficients given their state variables and the Jeffreys´ noninformative prior is assigned to the noise variance. Our signal reconstruction scheme is based on an EM iteration that aims at maximizing the posterior distribution of the signal and its state variables given the noise variance. We construct the missing data for the EM iteration so that the complete-data posterior distribution corresponds to a hidden Marcov tree (HMT) probabilistic graphical model that contains no loops and implement its maximization (M) step via a max-product algorithm. This EM algorithm estimates the vector of state variables as well as solves iteratively a linear system of equations to obtain the corresponding signal estimate. We select the noise variance so that the corresponding estimated signal and state variables obtained upon convergence of the EM iteration have the largest marginal posterior distribution. We compare the proposed and existing state-of-the-art reconstruction methods via signal and image reconstruction experiments.
Keywords :
Gaussian noise; compressed sensing; expectation-maximisation algorithm; hidden Markov models; signal reconstruction; trees (mathematics); white noise; Bayesian EM algorithm; Bayesian expectation-maximization algorithm; EM iteration; Gaussian priors; HMT probabilistic graphical model; Jeffreys noninformative prior; Markov-tree sparse signal reconstruction; belief propagation; binary state variables; complete-data posterior distribution; compressive samples; hidden Marcov tree probabilistic graphical model; image reconstruction; large-magnitude component; marginal posterior distribution; max-product EM algorithm; noise variance; posterior distribution maximization; probabilistic dependence structure; regression-coefficient vector; signal coefficient; small-magnitude component; state variable vector estimation; state variables; zero-mean white Gaussian noise; Decision support systems; Belief propagation; compressed sensing; expectation-maximization algorithms; hidden Markov models; signal reconstruction;
Journal_Title :
Signal Processing, IEEE Transactions on
DOI :
10.1109/TSP.2013.2277833