Bayesian mixture of gaussian processes for data association problem

Abstract

We address the data association problem and propose a Bayesian approach based on a mixture of Gaussian Processes (GPs) having two key components, the assignment probabilities and the GPs. In the proposed approach, the two key components are simultaneously updated according to observations through an efficient Expectation-Maximization (EM) algorithm that we develop. The proposed approach is thus more adaptive to the observations than the existing approaches for data association. To validate the performance of the proposed approach, we provide experimental results with real data sets as well as two synthetic data sets. We also provide a theoretical analysis to show the effectiveness of the Bayesian update.