Toeplitz-structured measurement for compressed channel feedback in FDD massive MIMO system

By exploiting the correlations of the massive MIMO channel, Compressed Sensing (CS) has been applied to develop efficient channel feedback scheme. The existing CS-based channel feedback schemes, however, adopt Gaussian random matrices to compress the channel, which will impose an unreachable memory and computation requirement on user equipment (UE). In this paper, a new Toeplitz-structured measurement matrix is proposed to perform efficient channels compression in massive MIMO system. Instead of containing entries of are independent realizations of random variables with certain distributions, Toeplitz matrices have constant diagonals, leading a significant reduction of UE requirement for storing and computation. Based on such matrices, we introduce novel feedback mechanism improving further the feedback efficiency for frequency division duplex (FDD) massive MIMO systems. Simulation results show that the Toeplitz-structured matrices perform comparably with Gaussian random matrices while requiring less independent random variables and computation complexity.