A nonlinear quantization decision approach for digital communication in channels with impulse noise

Turbo codes are effective for detection in channels with white Gaussian noise but not with some types of impulse noise because the probability of error becomes time dependent. Then the standard linear Hamming distance between code words is not representative of the error probability. This paper presents an alternative approach that uses the norm distance of the Voronoi partition as a decision function. A second order polynomial is used to estimate this function. The receiver's decision is based on the codeword that minimizes the second order norm function. The approach was simulated in a parallel CDMA model in order to use the information provided by cross-correlation between parallel data streams. A strong correlation between the linear square and cross-distances was found. This suggests that the coefficients for the proposed non-linear norm function could be estimated from pilot symbols and then used to decode the data.