Massive Connectivity Over MIMO-OFDM: Joint Activity Detection and Channel Estimation With Frequency Selectivity Compensation

In this paper, we study how to efficiently and reliably detect active devices estimate their channels in a multiple-input multiple-output orthogonal frequency-division multiplexing (OFDM) based grant-free non-orthogonal multiple access system enable massive machine-type communication (mMTC). First, by exploiting the correlation of channel frequency responses across OFDM subcarriers, propose block-wise linear model. Specifically, continuous subcarriers are divided into several sub-blocks function with only two variables (mean slope) is used approximate frequency-selective each sub-block. This significantly reduces number be determined estimation, sub-block can adjusted compensate frequency-selectivity. Second, formulate joint device detection estimation as Bayesian inference problem. By block-sparsity matrix, an efficient turbo message passing algorithm solve We then develop state evolution predict performance algorithm. further incorporate machine learning approaches learn unknown model parameters. Numerical results demonstrate superior proposed over state-of-the-art algorithms.