Skip to Main Content
This paper reports the results of a study of adaptive communication filtering. It is assumed that either the noise or the signal is completely unknown and only the spectrum or autocorrelation function of the other is known. it is desired to filter according to a mean-square error criteria. Several adaptive filters were investigated, and the two found practical are described here in detail. Both may be realized pnysically by the use of tapped delay lines, or they may be constructed as part of a digital computer program. On the computer they are capable of real-time filtering of many types of signals. Measurements taken of signal and noise give the information needed to adJust-- ia a straightforward manner--the filter parameters. The parameters will converge with time to give the optimum filter. Also described is a technique external to the adaptive filter for greatly increasing the useable frequency range when filtering a narrow-band signal.