It Is now possible with two knovra signals of different spectral 

 characteristics to study the response of the province picker as It 

 crosses the boundary between two such signals (provinces). Since it is 

 also necessary to detect changes in total RMS energy without a change of 

 slope, each signal can be multiplied by a constant using a corollary of 

 the addition theorem of Fourier transforms (see Bracewell, 1965). 



Figure B-6 illustrates an artificially generated random signal 

 which consists of four distinct provinces. The first half of the signal 

 is composed of "white noise" and the second half of 1/s noise generated 

 via the random walk technique. Notice the rapid change in the slope 

 parameter from to -1 at the boundary, which was easily detected. 



Within each half of the signal, the provinces are again divided 

 with the second halves (second and fourth quarters) representing a dou- 

 bling of the first halves (first and third quarters). Notice the 

 obvious change of RMS energy, although the slope parameter is unaf- 

 fected, illustrating the importance of detecting on two uncorrelated 

 parameters. Several false alarms are observed on the derivative detec- 

 tor, but the province boundaries derived from contouring (represented by 

 straight lines above the RMS energy profile) correspond to the known 

 boundaries in the signal. Notice that provinces one and four show the 

 same RMS energy level, but are delineated by their differences in spec- 

 tral slope. 



As stated earlier, the automatic detection is an aid to province 

 boundary recognition, but in some cases, human Intervention is needed to 

 resolve inconsistencies. Also, the setting of contour interval or slope 



145 



