Iterative Learning for Optimized Compressive Measurements in Massive MIMO Systems

Massive multiple-input multiple-output (MIMO) is a key enabling technology for next-generation wireless communications and future radar systems. The high number of antennas in a massive MIMO system makes it difficult to process array signals with separate radio frequency chains. We have previously developed a compressive measurement technique that substantially reduces the required number of front-end circuits and the overall processing complexity. By utilizing the a priori distribution of signal directions, the compressive measurement matrix is optimized by maximizing the mutual information between the compressive measurement output and the signal directions-of-arrival (DOAs). In practice, however, the prior distribution of the signal DOAs may not be available, thereby limiting the applications of this technique. In this paper, we develop an iterative learning scheme that uses the estimated DOA spectrum as the prior for the next iteration. By removing the hurdle of requiring the prior information of signal arrivals, the proposed technique makes optimized compressive measurement feasible for broader applications in practice.