AN INVESTIGATION OF A CORRELATION COVARIANCE METHOD OF SIGNAL EXTRACTION

A harmonic correlation technique has been used effectively to isolate a primary signal from two independent data sets each of which is conta minated by the presence of interfering signals. Each of the two data s ets is decomposed by a harmonic analysis, and then the correlation coe fficients and amplitude ratios between the two decompositions are comp uted fbr each harmonic. Only harmonics which satisfy predetermined sel ection criteria are retained. Those criteria are in the form of requir ing the correlation coefficient to exceed some magnitude, say, rho(0), and the amplitude ratio not to exceed some value, say, R(0). The resu lting pair of retained signals are then typically highly correlated. I t is suggested that appropriate values of rho(0) lie between 0.25 and 0.6 and of R(0) between 3.0 and 6.0, depending on the relative amplitu des of the primary and interfering signals. Critical examination of th e technique shows that under many conditions it can be an effective to ol for eliminating signals which are not a good approximation to the p rimary signal. This conclusion is confirmed by simulated examples of s atellite magnetic anomaly data with interfering fields from ionospheri c currents.