Mixed Generalized Gaussian Noise Channels: Capacity, Decoding, and Performance

In many different fields of communications such as sonar applications and Compressed Sensing, a growing interest in non-Gaussian channel models has been observed. One model of high interest is the so-called generalized Gaussian noise, which is often used to model impulsive events. In this paper, a survey of the properties of (mixed) additive white generalized Gaussian noise channels is given. The log-likelihood ratios are derived and connections to conventional LLR calculation are pointed out. We also discuss the capacity and give an upper and a lower bound for it. It is shown that the gain achievable by channel coding strongly depends on the parameters of the generalized Gaussian noise.