Compressed CSI Feedback With Learned Measurement Matrix for mmWave Massive MIMO

• DOI: 10.48550/arxiv.1903.02127
• Publication Date: 2019-01-01
• Authors: Pengxia Wu; Zichuan Liu; Julian Cheng
• Keywords: Matrix (chemical analysis)

A major challenge to implement the compressed sensing method for channel state information (CSI) acquisition lies in design of a well-performed measurement matrix reduce dimension sparse vectors. The widely adopted randomized matrices drawn from Gaussian or Bernoulli distribution are not optimal. To tackle this problem, we propose fully data-driven approach optimize beamspace compression, and trains mathematically interpretable autoencoder constructed according iterative solution recovery. obtained can achieve near perfect CSI recovery with fewer measurements, thus feedback overhead be substantially reduced.