Practical Polar Code Construction Over Memoryless Impulse Noise Channels

Polar coding subject to memoryless impulse noise described by a Gaussian mixture model is investigated. In contrast with existing approaches resorting to latency-laced interleaving, this paper tackles statistical dependence between in-phase and quadrature components of impulse noise, attributable to employing higher order modulation schemes, with meticulous revision in successive cancellation decoding. Furthermore, forfeiting using the statistical knowledge of impulse noise, a density evolution-based code construction method is proposed. Challenges of the aggravated complexity-ensued from addressing the aforementioned statistical dependence-are subsequently overcome by an approximation approach. The block error rate performance for the proposed framework is investigated by computer simulations and the experimental results exhibit the markedly performance gain over the conventional construction method under certain circumstances.