280 Power SpectrumAnalysis

As the number of signal samples N increases, we have

lim
N
E[

P
XX
( f )] =

m=
r
xx
(m)e
j2fm
=P
XX
( f ) (10.27)
For a Gaussian random sequence, the variance of the periodogram can be obtained as
Var[

P
XX
( f )] =P
2
XX
( f )
_
1 +
_
sin 2fN
N sin 2f
_
2
_
(10.28)
As the length of a signal record N increases, the expectation of the periodogram converges to the power
spectrum P
XX
( f ) and the variance of

P
XX
( f ) converges to P
2
XX
( f ). Hence the periodogram is an unbiased
but not a consistent estimate.
The periodograms can be calculated froma DFTof the signal x(m), or froma DFTof the autocorrelation
estimates r
xx
(m). In addition, the signal from which the periodogram, or the autocorrelation samples, are
obtained can be segmented into overlapping blocks to result in a larger number of periodograms, which
can then be averaged. These methods and their effects on the variance of periodograms are considered
in the following.
10.4.2 Averaging Periodograms (Bartlett Method)
In this method, several periodograms, fromdifferent segments of a signal, are averaged in order to reduce
the variance of the periodogram. The Bartlett periodogramis obtained as the average of K periodograms as

P
B
XX
( f ) =
1
K
K

i=1

P
(i)
XX
( f ) (10.29)
where

P
(i)
XX
( f ) is the periodogram of the i
th
segment of the signal. The expectation of the Bartlett
periodogram

P
B
XX
( f ) is given by
E[

P
B
XX
( f )] =E[

P
(i)
XX
( f )]
=
N1

m=(N1)
_
1
|m|
N
_
r
xx
(m)e
j2fm
(10.30)
=
1
N
1/2
_
1/2
P
XX
(v)
_
sin ( f v)N
sin ( f v)
_
2
dv
where (sin fN/sin f )
2
/N is the frequency response of the triangular window 1 |m|/N. From Equa-
tion (10.30), the Bartlett periodogram is asymptotically unbiased. The variance of

P
B
XX
( f ) is 1/K of the
variance of the periodogram, and is given by
Var
_

P
B
XX
( f )
_
=
1
K
P
2
XX
( f )
_
1 +
_
sin 2fN
N sin 2f
_
2
_
(10.31)
10.4.3 Welch Method: Averaging Periodograms from Overlapped and Windowed
Segments
In this method, a signal x(m), of length M samples, is divided into K overlapping segments of length N,
and each segment is windowed prior to computing the periodogram. The i
th
segment is dened as
x
i
(m) =x(m+iD), m=0, . . . , N 1, i =0, . . . , K 1 (10.32)
Non-Parametric Power Spectrum Estimation 281
where D is the overlap. For half-overlap D=N/2, while D=N corresponds to no overlap. For the i
th
windowed segment, the periodogram is given by

P
(i)
XX
( f ) =
1
NU

N1

m=0
w(m)x
i
(m)e
j2fm

2
(10.33)
where w(m) is the window function and U is the power in the window function, given by
U =
1
N
N1

m=0
w
2
(m) (10.34)
The spectrumof a nite-length signal typically exhibits side-lobes due to discontinuities at the endpoints.
The window function w(m) alleviates the discontinuities and reduces the spread of the spectral energy
into the side-lobes of the spectrum. The Welch power spectrumis the average of K periodograms obtained
from overlapped and windowed segments of a signal:

P
W
XX
( f ) =
1
K
K1

i=0

P
(i)
XX
( f ) (10.35)
Using Equations (10.33) and (10.35), the expectation of

P
W
XX
( f ) can be obtained as
E[P
W
XX
( f )] =E[

P
(i)
XX
( f )]
=
1
NU
N1

n=0
N1

m=0
w(n)w(m)E[x
i
(m)x
i
(n)]e
j2f (nm)
=
1
NU
N1

n=0
N1

m=0
w(n)w(m)r
xx
(n m)e
j2f (nm)
=
1/2
_
1/2
P
XX
(v)W(v f )dv (10.36)
where
W( f ) =
1
NU

N1

m=0
w(m)e
j2fm

2
(10.37)
and the variance of the Welch estimate is given by
Var[

P
W
XX
( f )] =
1
K
2
K1

i=0
K1

j=0
E
_

P
(i)
XX
( f )

P
(j)
XX
( f )
_

_
E
_

P
W
XX
( f )
__
2
(10.38)
Welch has shown that for the case when there is no overlap, D=N,
Var[P
W
XX
( f )] =
Var[P
(i)
XX
( f )]
K
1

P
2
XX
( f )
K
1
(10.39)
and for half-overlap, D=N/2,
Var[

P
W
XX
( f )] =
9
8K
2
P
2
XX
( f ) (10.40)
282 Power SpectrumAnalysis
10.4.4 BlackmanTukey Method
In this method, an estimate of a signal power spectrum is obtained from the Fourier transform of the
windowed estimate of the autocorrelation function as

P
BT
XX
( f ) =
N1

m=(N1)
w(m) r
xx
(m)e
j2fm
(10.41)
For a signal of N samples, the number of samples available for estimation of the autocorrelation value at
the lag m, r
xx
(m), decrease as m approaches N. Therefore, for large m, the variance of the autocorrelation
estimate increases, and the estimate becomes less reliable. The window w(m) has the effect of down-
weighting the high variance coefcients at and around the end points. The mean of the BlackmanTukey
power spectrum estimate is
E[

P
BT
XX
( f )] =
N1

m=(N1)
E[ r
xx
(m)]w(m)e
j2fm
(10.42)
NowE[ r
xx
(m)] =r
xx
(m)w
B
(m), where w
B
(m) is the Bartlett, or triangular, window. Equation (10.42) may
be written as
E[

P
BT
XX
( f )] =
N1

m=(N1)
r
xx
(m)w
c
(m)e
j2fm
(10.43)
where w
c
(m) = w
B
(m)w(m). The right-hand side of Equation (10.43) can be written in terms of the
Fourier transform of the autocorrelation and the window functions as
E[

P
BT
XX
( f )] =
1/2
_
1/2
P
XX
()W
c
( f )d (10.44)
where W
c
( f ) is the Fourier transform of w
c
(m). The variance of the BlackmanTukey estimate is
given by
Var[

P
BT
XX
( f )]
U
N
P
2
XX
( f ) (10.45)
where U is the energy of the window w
c
(m).
10.4.5 Power Spectrum Estimation from Autocorrelation of Overlapped Segments
In the BlackmanTukey method, in calculating a correlation sequence of length N from a signal record
of length N, progressively fewer samples are admitted in estimation of r
xx
(m) as the lag m approaches
the signal length N. Hence, the variance of r
xx
(m), increases with the lag, m. This problem can be solved
by using a signal of length 2N samples for calculation of N correlation values. In a generalisation of
this method, the signal record x(m), of length M samples, is divided into a number K of overlapping
segments of length 2N. The i
th
segment is dened as
x
i
(m) =x(m+iD), m=0, 1, . . . , 2N 1 (10.46)
i =0, 1, . . . , K 1
where D is the overlap. For each segment of length 2N, the correlation function in the range of 0mN
is given by
r
xx
(m) =
1
N
N1

k=0
x
i
(k)x
i
(k +m), m=0, 1, . . . , N 1 (10.47)
In Equation (10.47), the estimate of each correlation value is obtained as the averaged sumof N products.