DCT and VQ based limited feedback in spatially-correlated massive MIMO systems

We consider the limited feedback problem in a multi-cell spatially-correlated massive MIMO frequency division duplexing (FDD) system. The very high dimension of massive MIMO channels makes the limited feedback schemes in traditional MIMO inefficient. This motivates us to explore a novel approach to reduce the feedback overhead in two stages: 1) The original channel vector is sparsified in the spatial frequency domain by the discrete cosine transform (DCT); 2) Since empirical normality tests imply the large DCT coefficients approximately satisfies Gaussian distribution, the large DCT coefficients can be further compressed by vector quantization (VQ). By this DCT and VQ based limited feedback approach, a very high dimensional channel vector can be indicated by only an index with the analyzable distortion. Numerical simulations also illustrate its effectiveness.