A Layered Detect-Compress-and-Forward Coding Scheme for the Relay Channel

We develop a practical layered detect-compress-and-forward (LDCF) coding framework for a three-node relay network. Assuming all transmissions to be mapped to Quadrature Amplitude Modulation (QAM) constellations, the LDCF framework at the relay utilizes maximum- likelihood detection, followed by a multi-level convolutional coding structure to achieve distributed joint source-channel coding of the detected indices. The layering is achieved by mapping the more important detection bit-planes to the more significant bit-planes of the QAM constellation, thus providing them a higher level of protection. As a result, the proposed strategy adapts to heterogeneous channel conditions by letting the destination extract as many CF cooperation layers as the channel conditions allow. At the destination, we propose an iterative message passing algorithm to achieve joint decoding of the relay's detection layers before recovering the original source information. Simulations results with off-the-shelf convolution codes indicate a performance gain of approximately 3 dB over the traditional unlayered CF strategy.