Bayesian conjugate analysis for multiple phase estimation

We propose a Bayesian conjugate framework for inferring multiple phases. The framework requires a generalisation of the von Mises distribution for multiple variables. The principal difficulty in the generalisation is the computation of the first order moment and the normalising constant which are essential for Bayesian inference. We propose two approaches, one based on a Bessel function expansion and the other based on a Markov Chain Monte Carlo technique using the Gibbs sampler. We then assess the performance of these two methods against variations in parameters of the generalised von Mises distribution.