Channel parameter estimation and localization based on structured sparse bayesian CP decomposition in mmWave massive MIMO-OFDM systems

In this paper, we propose a structured sparse Bayesian CANDECOMP/PARAFAC (SSBCP) algorithm for channel parameter estimation and localization in millimeter-wave (mmWave) massive multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems with receiver-side hardware impairments. Firstly, based on the physical mechanisms of receiver hardware impairments, the received signal containing both antenna-dependent sparse noise and additive Gaussian noise is constructed as a third-order parallel factor (PARAFAC) tensor. Secondly, by designing an equivalent hybrid precoding matrix, the original complex-valued tensor is transformed into a real-valued counterpart suitable for Bayesian processing. Thirdly, accurate estimation of the factor matrices is achieved through a structured sparse Bayesian tensor decomposition that incorporates binary latent variables to control the positions of sparse noise. Finally, the channel parameters are extracted from the estimated factor matrices and the localization is accomplished based on their geometric relationships. Simulation results show the proposed SSBCP algorithm outperforms existing algorithms across sparse noise ratios. Even under severe hardware distortion conditions, the proposed SSBCP algorithm maintains outstanding parameter estimation and localization performance in environments with multiple scattering paths.