Higher order statistics of communication signals via tensor algebra

Building blocks for the computation of moment tensors of communication signals are given. Some nonlinearities are approximated by their power series. In most cases a relationship that corresponds to the transformation of the underlying signal is given for the cumulants. It is found that care has to be taken when decimating cyclostationary signals that the subsampling period is not commensurate with a cyclostationary period. If it is, then the moments can only be interpreted as averages over the possible phases and should not be looked upon as measurable values.<<ETX>>