Optimized metric clipping decoder design for impulsive noise channels at high signal-to-noise ratios

In many practical communication systems, the channel is corrupted by non-Gaussian impulsive noise (IN). It introduces decoding metric mismatch for the traditional Euclidean metric decoders and limits system performance. The situation is worsen by the practical difficulty in accurately estimating the IN statistics. Recently, some metric clipping based decoders with a properly chosen clipping threshold has been shown to be very effective in mitigating the effect of IN, even without a precise knowledge of its statistics. However, we observe that such a clipping threshold is derived based on some assumptions which lead to an error floor in the bit error probability curve at high signal-to-noise ratio (SNR). In this work, a clipping threshold is derived by an optimization approach without exploiting the IN statistics. It is demonstrated by experiment that with our proposed clipping threshold, the optimized metric clipping decoder is able to perform close to the maximum likelihood decoding performance at high SNR under the Bernoulli Gaussian noise model with various parameters.