A Gamma Beta Mixture Model for Channel Multipath Components Clustering

In this paper, a Gamma Beta Mixture Model (GBMM) is proposed to cluster channel multipath components (MPCs) where the gamma distribution is utilized to fit the delay data and angle data is fitted with beta distribution. We utilize the expectation-maximization (EM) algorithm to get the optimal GBMM parameters. Specially, in the M step of the EM algorithm, the Newton-Raphson method is utilized to optimize the GBMM parameters since we could not get closed solutions. To verify the clustering effect, an outdoor-to-indoor (O2I) measurement activity at 3.5 GHz was conducted. Simulation results based on real channel measurement data indicate that, compared with Gaussian mixture model (GMM), GBMM has better clustering performance.