TIME-FREQUENCY TRANSFORMATIONS WITH
GENERALIZED MARGINALS FOR NON-STATIONARY SIGNAL ANALYSIS
IN INFORMATION-MEASUREMENT SYSTEMS
A. V. Merkusheva
For non-stationary signal analysis
in information-measurement systems (IMS), a set of time-frequency distributions
(TFD) with generalized marginals (GM) is considered, i.e. marginals beyond
the time or frequency domain. The TFD projections in the form of integrals
along one or several directions (at several different angles) on the time-frequency
plane are used as GMs. GM values are equal to the module square of the rotated
Fourier transform. The condition is given, on realization of which the Сohen-class
TFDs have GM. The well-known TFDs of Сohen-class (Choi-Williams, Page
TFDs) may be modified so that to have GM. From the point of view of signal
presentation, these modifications have the advantages over quadratic TFDs
having only ordinary marginals. Moreover, the TFDs with GM are more robust
in conditions when the IMS signal includes additive noise.