Periodicity estimation by hypothesis-directed search

Some methods are described for estimating the fundamental periodicity of additively combined quasi-periodic signals. These methods operate from measurements of the instantaneous amplitudes and frequencies of important sinusoidal components of the input signal. The methods share a similar computational structure, in which each component is allowed to assert a number of hypotheses as to possible fundamental periodicities to which it is related. Hypotheses from different components are combined in such a way as to reinforce common periodicities shared by several components. The methods discussed seem well suited to signals whose fundamental frequencies may be high, and vary over a wide range.