Shape coded modulation using neural networks

Shape coded modulation is a combined modulation and error-control coding technique. The performance of a neural network decoder for shape coded modulation is compared with a decoder based on a complex algorithm. Simulation results over a range of signal-to-noise ratios (SNRs), for rate 1/2 and rate 1/4 shape codes, show that a single neural network decoder trained at a high SNR, provides near optimum results over all SNRs.