Enhanced ensemble averaging using the time-frequency plane

The paper describes the initial development of an optimal filter, based on Wiener´s theory of filtering and prediction. The optimal filter works in an a posteriori manner, meaning that no statistics of either signal or noise processes are assumed. The design is tailored to cycle-stationary signals mixed with uncorrelated noise. The procedure is a potentially superior estimator to ensemble averaging. The setting of the filter is the time-frequency plane. This provides a convenient means of accommodating signal nonstationarity, a major problem in the estimation of many real-world signals. The method is generally applicable to any repetitive, noisy signal with a deterministic time-frequency structure. The optimal filter has been applied successfully to the high resolution electrocardiogram for the recovery of low-level cardiac signals