77 Estimates ofthe Power Spectrum
may or may not be from an AR process. This knowledge is often unknown. Neverthe-
Jess, we are modeling the data as if it were an AR data model. Therefore:
i) =-Daxe-w 7738]
——
x(0) = - 3 axl - 1) + 0(0) = &(0) +06) 7391
hich we illustrate in Figure 7.23,
‘The z-transform of [7.7.3.9] is
x@=-| Sax ‘}X@ +E) = RO) +E pza.0)
ich can be expressed as
X(z) = = 773.01
140 ae) 73121
jon [7.7.3.11] is implemented as a filter in Figure 7.24, This form is often called
Seward filter model, since E(2), the input excitation, is filtered by 1/A(z) to produce
i the LP data,
We can also write [7.7.3.10] in the following form:
E(@) =[1+ Sar4]x@) =a@ xia) 773.131
ahr)
Figure7.23 Linear prediction model