Compressed sensing for MIMO radar: A stochastic perspective

Compressed sensing for MIMO radar can potentially enhance spatial resolution and improve anti-jamming capability by virtue of multiple transmitter and receiver antennas, and at the same time reduces the number of samples needed by making use of the inherent sparsity property of most radar scenes. Existing work along this line adopts a deterministic model for the radar signals, which may not be effective to cope with fading propagation and signal correlation in practical scenarios. This paper takes a stochastic approach by modeling the target scenes as random processes that are possibly correlated. A new stochastic framework of compressed sensing for MIMO radar is developed for reconstructing useful statistics of the random target scenes using a small number of samples. The proposed approach directly extracts the useful statistics for estimation without reconstructing the random signals; as a result, it is computationally more efficient and requires a smaller number of samples than existing deterministic approach to compressed sensing.