Abstract :
An attempt is made to bring together the point of view which has governed the design of frequency filters and that which governs "optimum filters" (Wiener, Zadeh-Ragazzini), where the desired signal is in the presence of a stationary random noise whose autocorrelation function is given. After having defined the class of filters resulting from this rapprochement, the integralequation on which it depends is written down. Various possible statistical interpretations of this method of filtering are then given. A number of simple examples are treated. Finally, a comparison with the point of view of Zadeh and Ragazzini is developed with reference to a particular example.
