Using the EM algorithm to increase the number of signals estimable by 2-D parametric techniques

This paper provides an easy method for increasing the number of modes estimable by 2-D parameter estimation schemes which use a single snapshot of the data. For data of size M/sub 1//spl times/M/sub 2/ few of these techniques allow the number of modes to exceed M/sub 1/ and/or M/sub 2/. As these restrictions are not inherent to the model but the algorithms themselves, this problem is circumvented by treating the observed data as an incomplete version of a larger data set. An existing 2-D parameter estimation scheme is then used for the maximization (M) step of the expectation-maximization (EM) algorithm. In this way, one can increase the number of modes the scheme can estimate without changing the scheme itself.