Hybrid Convolutional Beamspace Method for mmWave MIMO Channel Estimation

Millimeter-wave (mmWave) MIMO channel estimation is studied. To reduce hardware cost, hybrid analog and digital processing is used. Hybrid convolutional beamspace (CBS) method is proposed for estimation of the channel. This method is especially attractive for large arrays, which have received more attention recently. In particular, a nonuniform scheme of CBS is proposed. The receiver combiner is a CBS filter followed by a nonuniform decimator, and the transmitter precoder is a nonuniform expander followed by a CBS filter. Although the analog precoder and analog combiner should have unit-modulus entries, it is shown that any CBS filter coefficients are realizable. The nonuniform decimation or expansion corresponds to antenna locations of a virtual sparse array, dilated by an integer factor. Thus, given a small number of RF chains, meaning low hardware complexity, a significant number of paths can still be estimated with difference coarray methods. More importantly, due to the dilation and sparse array structure, a larger coarray aperture is achieved, resulting in better estimation performance. The advantages of the proposed method are shown by simulations.