A novel method to obtain CSI based on Gaussian Mixture Model and Expectation Maximization

Abstract

Wireless channel is the key and fundamental part of wireless communication. How to get the accurate Channel State Information (CSI) is a very challenging and meaningful issue. In this paper, we analyze the relationship between the radio propagation environment and the channel state information (CSI) to get the statistical property of wireless channel and propose a layered channel model. The layered channel model consists of static channel information, dynamic channel information and disturbing channel information. Based on the layered channel model, we apply the Gaussian Mixture Model (GMM) and Expectation Maximization (EM) to obtain the static CSI from data of the indoor and outdoor channel measurement campaigns. With the static channel information, velocity of MS and the character of random scatters, we can get the full CSI of the wireless channel. As numerical results demonstrate, the proposed method can extract accurate static CSI from real repeated measurement data and based on the full CSI we have gotten, we can accomplish channel estimation with less pilot resource or even without pilot resource.