Bandwidth expansion analog joint source-channel coding with channel inversion and multiple receive antennas

We investigate the performance of bandwidth expansion analog joint source-channel coding (JSCC) over fading channels. Non-linear spiral-like curves and maximum likelihood detector with linear minimum mean square error estimator are considered. Under Rayleigh fading this scheme presents a gap to the theoretical limit which increases with the channel signal to noise ratio (CSNR), while the same scheme on additive white Gaussian noise (AWGN) channel presents much smaller gap. This is completely different from bandwidth compression, where the performance in both fading and AWGN channels is close to optimal. We present two strategies that can make bandwidth expansion analog JSCC more feasible: channel inversion and receive diversity. The first requires channel state information at the transmitter, while the second does not. Both strategies are shown to considerably reduce the gap to the theoretical limit.