Fastest-known near-ML decoding of golden codes

We present a reliability-based near-maximum-likelihood (near-ML) decoder of Golden code with significantly reduced average decoding complexity over the current state of the art We first pre-process the received signal via zero-forcing (ZF) filtering and compute symbol reliabilities at the filter output Reliable symbols are directly decoded and removed from the received signal. The remaining symbols are decoded by reduced-dimension ML or near-ML decoding. Computational studies included herein reveal extensive complexity savings when compared with state-of-the art ML and near-ML decoders for the Golden code. For instance, for the 16-QAM signal constellation at pre-detection signal-to-noise ratio of 28dB that corresponds to bit-error-rate of about 10^-4, the presented algorithm achieves more than 96.6% average-complexity savings, while maintaining indistinguishable to ML performance.