Sparse Channel Estimation in MM - Wave Hybrid MIMO systems

Abstract

Millimeter wave (mm-Wave) is emerging as one of the predominant 5G technology at high frequency. In this paper the channel estimation at mm-Wave is formulated as a sparse problem in which the hybrid multiple- input multiple-output (MIMO) precoders and combiners are used as the measurement matrices. The hybrid MIMO system judiciously partitions mm-Wave precoding-combining between analog and digital domain, due to high power consumption and cost of mixed signal devices. Exploiting the sparsity of mm-Wave channels, a (Compressed sensing) CS problem is formulated that estimates the angle of departure/arrival and gain of each corresponding path. Mm-Waves employs the directional beam forming which divides the angle of arrival and the angle of departure space intogrids. A dictionary is created where all the possible angles of arrivals of the received array response vectors corresponding to a particular resolution are placed. Given the RF beamforming matrix using discrete fourier transform (DFT matrix), baseband precoder-combiner matrix (assumed to be unitary), the orthogonal matching pursuit (OMP) algorithm, ORACLE-LS estimator performance is compared in terms of normalized mean square error (NMSE) on a virtual channel model in mm-Wave hybrid MIMO system. The MATLAB simulation results shows the advantage of low complexity OMP estimator which evaluates NMSE using fewer samples compared to ORACLE-LS estimator which requires full training samples, and it’s estimation error has shown to approach Cramer Rao lower bound.