Skip to Main Content
Abstract—This paper deals with theoretical and experimental aspects of estimating correlation functions of signals generated by nonstationary processes. The estimation process is based on approximation of the expectation or ensemble average definition of the correlation function. The assumption of stationarity regarding the signals under analysis need not be made. Emphasis is placed on off-line analysis, where immediate readout of a correlation-function estimate is not required. Hybrid or digital computation may be used for off-line analysis, resulting in drift-free, repeatable results. The estimation procedure is developed with the aid of filtering theory. Three different solutions for estimation are presented: two-dimensional filtering, one-dimensional filtering, and one-dimensional filtering with deficient a priori data. Experimental results obtained from the digital computer are presented for the latter solution.