Combined Source and Channel Coding for Channels with a Power Constraint and Multilevel Signaling

In this paper, a method for optimizing the index map between the output indices of a vector quantizer and the channel symbols for a channel with a power constraint and multilevel signaling is presented. The method is based on simulated annealing and makes use of dynamic scaling of the signal constellation during the optimization procedure in order to optimize for a fixed transmitted power. The method is applied to a memoryless Gaussian source and to the subband samples of a still image subband coder. The resulting system is shown to give a significant gain in terms of channel signal-to-noise ratio when compared to a reference system based on arbitrary index maps.