Turbo Code Performance in Heavy-Tailed Noise

Most of the research performed in the area of turbo codes has considered only Additive White Gaussian Noise (AWGN) channels. While some work has extended the coverage of channel noise to fading channels, thus far little research has been performed with heavy-tailed noise distributions. This paper analyzes the performance of the turbo decoder for a class of heavy-tailed noise distributions consisting of the Gaussian mixture, double exponential , and Cauchy distributions. The decoder is modi-ed to handle non-Gaussian noise models without significantly increasing complexity. We show that the performance relative to channel capacity is nearly constant over the class of distributions, roughly mirroring that on the Gaussian channel. Also, we explore decoder robustness over the class of heavy-tailed noise.