The recursive maximum likelihood estimation (RMLE) algorithm
conceived by Kay (see ibid., vol.ASSP-31, p.56, 1983) is extended to
complex data sets. The complex version requires the same level of
computation as that for real data. The original development was
restricted to the case of realm data. The purpose is to extend RMLE to
the more universal realm of complex data. The derivation is discussed.
It is argued, without direct proof, that the algorithm is stable in the
sense that the magnitude of the reflection coefficient at each step is
less than unity