| Abstract: | In the case of simple time series it is shown that prediction operators can be considered as deconvolution operators which are easily obtained. These operators possess the special feature of having a minimum phase, and their spectrum modulus represents, with a good dynamic range, the reciprocal of the square root of the modulus of the original autocorrelation spectrum. A generalization of the simple time series theory to the functions of two independent variables is possible in as much as, on a given section, the application of the multiple time series method enables the function of time and distance variables to have well-defined statistical properties; it is necessary, in particular, that the processes involved are stationary with respect to the two independent variables. In the case of multiple time series the application of the Prediction Theory permits greater uniformity of the traces because it enhances the events which show a good correlation between traces and, on the other hand, tends to minimize the random events which cannot be correlated between traces. The signal-to-noise ratio is thus increased to a great extent. |
