Hybrid Digital-Analog Coding of Memoryless Gaussian Sources over AWGN Channels with Bandwidth Compression

We present a low-complexity and low-delay joint source-channel coding method for bandwidth compression using a hybrid digital-analog (HDA) coding/modulation system based on the recent work in Skoglund, M et al., (2005), Analytical optimal distortion expressions (under both matched and mismatched channel conditions) are obtained for the proposed HDA system with a linear analog part for a memoryless Gaussian source and additive white Gaussian noise (AWGN) channel under the mean squared error distortion measure. We consider two HDA coding schemes, both of which employ a vector quantizer cascaded with binary phase-shift keying (BPSK) modulation in the digital part, but differ in that they use linear (resp. non-linear) coding with pulse amplitude modulation (PAM) in the analog part. We derive an optimal power allocation scheme for the system with linear analog coding and present performance comparisons with purely analog and purely digital systems. Simulation results show that, under linear analog coding, the proposed scheme outperforms the medium to high channel signal-to-noise ratios (CSNRs). Furthermore, the performance of the HDA scheme with the linear analog part is within 1 dB of the optimal distortion bound for the mismatched HDA system for high CSNRs; for the scheme with non-linear analog coding, the performance can be improved at high CSNRs