Bandwidth Compression for a Bivariate Gaussian Source with Shannon-Kotel’nikov Mappings

The analog transmission of a bivariate Gaussian source over a white Gaussian channel with 2:1 bandwidth compression is addressed in this paper. We propose two nonlinear coding structures for bandwidth compression based on the Shannon-Kotel’nikov (S-K) mappings. For each coding structure, the closed-form expression of distortion at any given channel SNR is derived and the optimal energy scaling for each source is computed. We perform theoretical deduction and numerical evaluations to compare the performances of the proposed schemes with a baseline linear mapping scheme. Results show that nonlinear schemes are superior to the linear scheme when the channel condition is good and the difference between the two sources is not too large. Among the proposed nonlinear schemes, the design to combine the two sources for S-K mapping with the optimal energy scaling yields the best performance.