Distributed analog linear coding of correlated Gaussian sources over multiple access channels

We study the performance of distributed analog linear coding of correlated multivariate Gaussian sources over the multiple access channel. Assuming the correlation structure is known both at the encoder and the decoder, optimal linear encoding and decoding is introduced. A general performance analysis is presented for both random and optimal linear encoders and compared to a lower bound on the achievable theoretical limit. Simulation results show the agreement between the theoretical analysis and the practical implementation.