Analysis of the accuracy and sensitivity of algorithms for the processing of multidimensional signals and noise

An analysis of algorithms for the processing of multidimensional signals and noise shows that algorithms that take into account mutual cumulants (the cumulant algorithms) are superior to algorithms that do not take into acount mutual cumulants (noncumulant algorithms) in terms of estimation accuracy and sensitivity to deviations of model parameters from the real signal-noise situation. The cumulant algorithms, however, are significantly more complex than the noncumulant algorithms and require the use of faster and more costly processors. Therefore the cumulant algorithms should be used only when a high accuracy or sensitivity is required; in other cases the noncumulant algorithms are preferable.