Novel semi-blind estimation for turbo decoding in impulsive noise channel

International audienceIn order to calculate the branches metric in the maximum a posteriori algorithm of turbo decoder, it is mandatory to know the values of parameters of the noise contaminating the transmitted signal. In the case of a generalized Gaussian distribution impulsive noise, it is very difficult to estimate the shape parameter, because the noise is inseparable from transmitted signal at turbo decoder reception. Until now, few researches about shape parameter estimation for an impulsive noise on turbo codes have been presented, and existing estimation methods use only the high order statistics (HOS). In this paper, we propose a novel semi-blind method, that does not use the HOS, to estimate the shape parameter from only the received signal in the turbo decoder. This method is based on fractional lower order statistics and the probability that the received signal is the same sign as the transmitted signal modulated with BPSK. The results, in terms of root mean square error, show the advantage of our method over other methods using HOS in the case of impulsive noise