Scribd
Upload
131 views10 pages

Non-Parametric Methods: Power Spectrum Estimation

This document discusses non-parametric power spectrum estimation methods. It introduces applications of power spectrum estimation like feature extraction. The periodogram method is presented as a basic approach that uses the discrete Fourier transform to estimate the power spectrum without assuming an underlying data model. Improvements to the periodogram method are discussed, including modifying the window function and averaging periodograms of subsets of the data to reduce variance. The document evaluates power spectrum estimation methods based on their resolution, bias, and variance as the amount of data increases.

Uploaded by

hkajai
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
131 views10 pages

Non-Parametric Methods: Power Spectrum Estimation

This document discusses non-parametric power spectrum estimation methods. It introduces applications of power spectrum estimation like feature extraction. The periodogram method is presented as a basic approach that uses the discrete Fourier transform to estimate the power spectrum without assuming an underlying data model. Improvements to the periodogram method are discussed, including modifying the window function and averaging periodograms of subsets of the data to reduce variance. The document evaluates power spectrum estimation methods based on their resolution, bias, and variance as the amount of data increases.

Uploaded by

hkajai
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 10

1

Non-Parametric
Power Spectrum Estimation
Methods

Eric Hui
SYDE 770 Course Project
November 28, 2002

2
Introduction
Applications of Power Spectrum Estimation
(PSE):
Wiener Filter
Feature Extraction

Non-parametric PSE does NOT assume any
data-generating process or model (e.g.
autoregressive model).

3
Motivation
Ideal autocorrelation:


Actual autocorrelation:


Limited (finite length of) data due to:
Availability of data
Assumption of stationary

)
`

+
+
=

=

N
N n
N
x
n x k n x
N
k r ) ( ) (
1 2
1
lim ) (


=
+ =
k N
n
x
n x k n x
N
k r
1
0
) ( ) (
1
) (
4
Periodogram Method
) ( ) (
1

) ( ) (
1

) ( ) (
1
) (
1
0
k x k x
N
n x k n x
N
n x k n x
N
k r
N N
n
N N
k N
n
x
- =
+ =
+ =

=

=
n
x(n)
N 0
2
) (
1
) (
e e j
N
j
per
e X
N
e P =
D
T
F
T

redefined
as
n
x
N
(n)
N 0
5
Periodogram Method
{ }
{ }
) ( ) (
1

) ( ) (
1

) ( ) (
1
) (
1
0
1
0
1
0
k r
N
k N
k r
N
n x k n x E
N
n x k n x
N
E k r E
x
k N
n
x
k N
n
k N
n
x

= =
+ =
)
`

+ =



=

=

=
{ } ) (
) sin(
) sin( 1
2
1
) (

2
2
1
2
1
e e
e
e
t
j
x
j
per
e P
N
N
e P E -
|
|
.
|

\
|
=
D
T
F
T

N 0 -N
k
D
T
F
T

N
k N
k w

=
A
) (
0
2
2
1
2
1
) sin(
) sin( 1
) (
|
|
.
|

\
|
=
A
e
e
e
N
N
e W
j
k
6
Good Method?
Necessary conditions for mean-square
convergence:
Asymptotically Unbiased


Zero Variance
{ } ) ( ) (

lim
jw jw
N
e P e P E =

{ } 0 ) (

lim =

jw
N
e P Var
k
PSD
k
PSD
as N
k
as N
k
PSD PSD
7
Evaluation of Methods
Resolution
How much blurring effect is there on the power
spectrum?
Bias (Asymptotic)
Does the estimation approach the true value with
more data (i.e. as N increases)?
Variance
Does the amount of deviation from the true value
depend on the data length (i.e. N)?
k
PSD
k
PSD
as N
k
as N
k
PSD PSD
k
PSD
8
Different PSE Methods
Periodogram Method
Apply rectangular window to x(n) to get x
N
(n).
Modified Periodogram Method
Apply non-rectangular window to x(n) to get x
N
(n).
Bartletts Method
Average the Periodogram estimate of non-overlapping
sub-intervals of x(n).
Welchs Method
Average the Modified Periodogram estimate of overlapping
sub-intervals of x(n).
Blackman-Turkey Method
Apply non-triangular window to r(x).

k
N 0 -N
D
T
F
T

N
k N
k w

=
A
) (
k
0
2
2
1
2
1
) sin(
) sin( 1
) (
|
|
.
|

\
|
=
A
e
e
e
N
N
e W
j
k
as N
k
PSD PSD
9
Application: Feature Extraction
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
x 10
4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-250
-200
-150
-100
-50
0
50
Linearized PSD Slope (Horizontal)
PSD
l
i
n
e
a
r
i
z
e

repeat for whole image
10
Questions or Comments?

You might also like