AbstractWireless communications in high‐mobility situations face with doubly selective (DS) fading. In addition, many practical wireless channels exhibit a sparse multipath structure. This paper investigates the DS multiuser multiple‐input–multiple‐output (MIMO) channel estimation via compressed sensing. The DS channel is characterized by a general sparse basis expansion model (BEM). This model reduces the channel estimation complexity as the channel estimation problem is reduced to estimating the BEM coefficients instead of numerous channel parameters. In addition, a pilot pattern is proposed for multiuser MIMO orthogonal frequency division multiplexing systems, which includes guard and silent pilots to reduce intercarrier interference and interuser interference, respectively. Then, two beyond‐sparsity features (ie, block and joint sparsity) of the BEM coefficients are exploited to propose a beyond‐sparsity–based channel estimation method. Moreover, a pilot design algorithm is proposed to optimize the pilot positions and enhance the estimator performance. Simulation results demonstrate that the proposed method significantly improves the estimation accuracy over the existing approaches, with a much smaller pilot overhead.