On the Performance of SVD-Based Channel Estimations in Large-Scale Multi-Cell Multiuser MIMO Systems

This paper introduces the traditional channel estimation based on singular value decomposition (SVD). Based on the SVD algorithm, the channel matrix composed of the maximum KL singular vector is obtained by SVD decomposition of the received signal covariance matrix. At this time, the estimation matrix has a certain ambiguity, and then the LS algorithm of the short pilot sequence is used to accurately estimate the channel matrix. This channel estimation scheme is unbiased in the ideal case, but in the actual implementation, the average approximation of the covariance matrix of the received signal is required, which leads to the deterioration of the actual channel matrix estimation, which is quite different from the theoretical value. Therefore, we derive the closed mean square error (MSE) of the channel estimation that characterizes the above error sources, so as to deeply understand the influence of the above error sources on the estimation algorithm. And compare with Cramer-Rao bound(CRB). The data results verify this analysis.