Joint User Activity Detection and Channel Estimation in MC-GFMA Systems by Block Sparse Bayesian Learning With Threshold Optimization

Future wireless communications are expected to support massive connectivity in various applications, such as massive Machine-Type Communications (mMTC) and different types of IoT networks, where many applications have the data traffic of sporadic nature. To support these kinds of applications, grant free multiple-access (GFMA) has been recognized to be more efficient than the conventional granted multiple access (GMA). However, due to sporadic transmission, GFMA faces the main challenges of User Activity Detection (UAD) and Channel Estimation (CE). To meet these challenges, in this paper, a multicarrier GFMA (MC-GFMA) system is introduced for supporting massive connectivity. A block-sparse signal model is derived, where the Expectation Maximization assisted Block Sparse Bayesian Learning (EM-BSBL) algorithm is employed to solve the joint UAD and CE problem. Furthermore, to augment the performance of EM-BSBL algorithm in GFMA systems, the statistical properties of the activity weights generated by EM-BSBL algorithm are investigated, showing that the activity weights follow closely the Gamma distribution. Then, using the Gamma modelling of the activity weights, the Neyman-Pearson (NP) method is considered for optimizing the threshold used for decision making in the EM-BSBL algorithm. Finally, the performance of GFMA systems is comprehensively studied by numerical simulations. Our results and analysis demonstrate that MC-GFMA is a feasible signalling scheme for supporting a massive number of users transmitting sporadic information. With the aid of the EM-BSBL algorithm enhanced by the NP-assisted threshold optimization, MC-GFMA is robust for operation in the communications environments where active users are random and the number of them is highly dynamic.