Optimizing signal constellations for the output of a quantizer
M.G. Perkins · 2003
A technique is presented for optimizing signal constellations for the output of a quantizer. The quantizer outputs are assumed to occur with unequal probabilities, and the mean-square error introduced by the channel is used as the performance criterion. A memoryless Gaussian channel and a peak power constraint are assumed. Transmitter power gains of 1 dB-2 dB were demonstrated when the technique was used to optimize signal constellations for the output of a DPCM (differential pulse code modulation) image coder.>