BAND PASS 10 



BAND PASS 9 



♦# ■* " ♦» ' #♦♦!♦ # »< » » 'iN N t<iiH |«» 'Ik 



BAND PASS 5 



*i ff^ \^ \i ^ii WH' *fi i ^ i 'i ^^^ 



INPUT SIGNAL 



Figure B-5 Example of province picker output with "white noise" input 

 illustrated in Figure B-4. The input signal is shown at the 

 bottom, above which are the output of ten band-pass filters 

 convolved with the data. The lowest signal (Band Pass 1) is 

 the lowest frequency pass, while the highest (Band Pass 10) 

 is the highest frequency pass. Above each band-passed sig- 

 nal is the energy envelope calculated by the Hilbert Trans- 

 form method. Notice the large variability in energy for each 

 band despite the constant amplitude input. At the top, the 

 estimated spectral exponent (slope of log-transformed spec- 

 tra) and the band-limited RMS energy calculated along the 

 profile are plotted. Standard deviation of the slope par- 

 ameter is .2. The algorithm was designed to minimize this 

 "natural" variability. 



144 



