LMM based blind signal separation with hybrid sampling

Laplace Mixture Model (LMM) is used to characterize the distribution of observed data. The stationary distribution of the model parameters is obtained by the hybrid of Gibbs and Metropolis-Hastings sampling, by which the mixing matrix is further estimated. With the mixing matrix estimated, we separate the sources successfully. Simulation results show that the method proposed is very robust to initial values and performs better than traditional methods, at the same time the method does not need large calculation amount.