Variational Bayesian PARAFAC decomposition for Multidimensional Harmonic Retrieval

High resolution parameters estimation for Multidimensional Harmonic Retrieval problem is required in a variety of applications including radar, sonar, and communication, etc.. Recent approaches based on deterministic tensor decomposition show promising results. However, these methods raise difficulties to estimate the unknown number of targets. In this paper, we address this problem through reformatting it into a Bayesian framework. Since exact Bayesian estimation of the unknown parameters is intractable, an approximation scheme based on variational principle is developed. The significant features of this approach are that the unknown number of targets are efficiently estimated as a part of Bayesian inference process and moreover, it provides high estimation performance. Experimental results demonstrate the effectiveness of the proposed method.