Scribd
Upload
67 views

Spectrum Estimation: Presentation by Dr. K.Muthumeenakshi Asso - Prof / ECE SSN College of Engineering

The document discusses spectrum estimation methods. It describes the periodogram estimator and its properties. The periodogram is the Fourier transform of the data and estimates the power spectrum. It is a biased but consistent estimator. Its variance decreases as the number of samples increases. Examples are provided to illustrate the periodogram of simulated white noise and sinusoidal signals.

Uploaded by

dolly
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
67 views

Spectrum Estimation: Presentation by Dr. K.Muthumeenakshi Asso - Prof / ECE SSN College of Engineering

The document discusses spectrum estimation methods. It describes the periodogram estimator and its properties. The periodogram is the Fourier transform of the data and estimates the power spectrum. It is a biased but consistent estimator. Its variance decreases as the number of samples increases. Examples are provided to illustrate the periodogram of simulated white noise and sinusoidal signals.

Uploaded by

dolly
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 51

Spectrum Estimation

Presentation by
Dr. K.Muthumeenakshi Asso.Prof / ECE
SSN College of Engineering
Spectral Estimation
• Non Parametric Methods
– Periodogram method
– Modified periodogram method
– Bartlett’s method
– Welch method
– Blackmann Tukey method
• Parametric Methods
– ARMA model
– AR model
– MA model
Spectral Estimation
• Non parametric method /Classical method
– Find autocorrelation from the data set
– Use Wiener Khinchine theorem to find its power
spectrum
• Parametric / Non classical method
– Use of a model to estimate the power spectrum
– ARMA
– AR
– MA
Performance of an Estimator
• Based on
– Bias
– Variance
Performance of an Estimator
• Based on
– Bias
– Variance
• Bias
Bias is defined as,

Estimator is said to be unbiased if,

Estimator is said to be asymptotically unbiased if,


Performance of an Estimator
• Variance:
If the variance goes to zero as N -> , then the
estimator is said to be consistent.
Performance of an Estimator
• Variance:
If the variance goes to zero as N -> , then the
estimator is said to be consistent.

• Another form of convergence: Mean square sense


Periodogram Estimator
• The power spectrum of a WSS process is the Fourier transform
of the autocorrelation sequence,

• Spectrum estimation – autocorrelation estimation problem.


Periodogram Estimator
• The power spectrum of a WSS process is the Fourier transform
of the autocorrelation sequence,

• Spectrum estimation – autocorrelation estimation problem.


• For an autocorrelation ergodic process with large data records,
Periodogram Estimator
• For finite data,
Periodogram Estimator
• For finite data,

• To exclude data that falls outside [0, N-1],


Periodogram Estimator
• Using conjugate symmetry of autocorrelation,

• is defined as the estimate of the power spectrum and


called periodogram.
Periodogram Estimator
• To express the periodogram in terms of x[n],
Periodogram Estimator
• To express the periodogram in terms of x[n],
Periodogram Estimator
• To express the periodogram in terms of x[n],
Periodogram Estimator
• To express the periodogram in terms of x[n],
Periodogram Estimator
• To express the periodogram in terms of x[n],

where,
Periodogram Estimator
Thus the periodogram is proportional to the squared magnitude of the DTFT of
xN(n) and may be computed using a DFT as follows:
Example 1
white noise process
amplitude 5

-5
0 5 10 15 20 25 30 35 40 45 50
samples
autocorrelation
100
amplitude

50

-50
-50 -40 -30 -20 -10 0 10 20 30 40 50
samples
power spectrum
magnitude(dB)

-2

-4

0 0.5 1 1.5 2 2.5 3 3.5


frequency(units of pi)
Example 1
white noise process
amplitude 5

-5
0 20 40 60 80 100 120 140 160 180 200
samples
autocorrelation
200
amplitude

100

-100
-200 -150 -100 -50 0 50 100 150 200
samples
power spectrum
2
magnitude(dB)

-2

-4

0 0.5 1 1.5 2 2.5 3 3.5


frequency(units of pi)
Example 2
sine (200Hz) + white noise process
amplitude 2

-2
0 10 20 30 40 50 60 70 80 90 100
samples
autocorrelation
100
amplitude

50

-50
-100 -80 -60 -40 -20 0 20 40 60 80 100
samples
power spectrum
0.06
magnitude(dB)

0.04

0.02

0
0 50 100 150 200 250 300 350 400 450 500
frequency
Example 2
sine (200Hz) + white noise process
amplitude 2

-2
0 50 100 150 200 250 300 350 400 450 500
samples
autocorrelation
500
amplitude

-500
-500 -400 -300 -200 -100 0 100 200 300 400 500
samples
power spectrum
0.2
magnitude(dB)

0.1

0
0 50 100 150 200 250 300 350 400 450 500
frequency
Performance of the Periodogram
• Mean square convergence,
Performance of the Periodogram
• Mean square convergence,

which implies,
Periodogram bias
• To compute the bias, we first evaluate the
expected value of
Periodogram bias
• To compute the bias, we first evaluate the
expected value of
Periodogram bias
• To compute the bias, we first evaluate the
expected value of
Periodogram bias
• Using the conjugate symmetry of
Periodogram bias
• Using the conjugate symmetry of
Periodogram bias
• Using the conjugate symmetry of

Therefore, is a biased estimate.


Periodogram bias
• The expected value of the periodogram is,
Periodogram bias
• The expected value of the periodogram is,
Periodogram bias
• The expected value of the periodogram is,
Periodogram bias
• The expected value of the periodogram is,
Periodogram bias
• The FT of the Bartlett’s window is,
Periodogram bias
• The FT of the Bartlett’s window is,
Periodogram bias
• The FT of the Bartlett’s window is,
Variance of the Periodogram
• Variance of the periodogram depends on the fourth
order moments, so we consider x[n] as a white noise
process,
Variance of the Periodogram
• Variance of the periodogram depends on the fourth
order moments, so we consider x[n] as a white noise
process,
Variance of the Periodogram
• Variance of the periodogram depends on the fourth
order moments, so we consider x[n] as a white noise
process,
Variance of the Periodogram
• The second order moment of the periodogram,
Variance of the Periodogram
• The second order moment of the periodogram,

• Using moment factoring theorem,


Variance of the Periodogram
• The first term becomes,
Variance of the Periodogram
• The first term becomes,

• The second term,


Variance of the Periodogram
• The first term becomes,

• The second term,


Variance of the Periodogram
• The first term becomes,

• The second term,


Variance of the Periodogram
• Combining the first and second terms,
Variance of the Periodogram
• Combining the first and second terms,

• We have the covariance of the periodogram,


Variance of the Periodogram
• As,
Variance of the Periodogram
• As,

• Setting
Performance of the Periodogram

You might also like