Application of joint source-channel decoding to impulsive noise environments

In recent years, joint source-channel (JSC) decoding has been in the focus of several research activities. In particular, the performance of JSC decoding in AWGN (additive white Gaussian noise) scenarios has been addressed. However, communication systems impaired by impulsive noise, like e.g. power line and digital subscriber line systems, but also mobile radio systems, have not yet been considered. The authors hence focus on examining joint source-channel (JSC) decoding in impulsive noise environments. We address the inclusion of a-priori information into the metric processing at the bit level processor, which generates the extrinsic information to be fed back to the FEC (forward error correcting) decoder. Four possible approaches to this a-priori metric processing, avoiding potential numerical instabilities, are illustrated and compared w.r.t. the achievable performance in both static as well as Rayleigh fading channel conditions. The authors demonstrate that a bit-level ML (maximum likelihood) based approach is well suited. After eight decoding iterations, performance improvements of about 2dB at an SER (symbol error ratio) of 10^-3 could be obtained.