Welch PSD Estimator Using Winows: SL - No. Title

WELCH PSD ESTIMATOR USING WINOWS
CONTENTS
Sl.no. Title
1. Abstract
2. Introduction
3. Power spectral density
4. Welch method
5. Result
6. conclusion
ABSTRACT
A good resolution in PSD may be achieved by using optimum size of data sample. In this
paper PSE has been performed for variable data length using Rectangular and Hamming
window with Welch Method. Welch method is nonparametric method that include the
periodograms that have the advantage of possible implementation using the fast Fourier
Transform. The periodograms technique based on Welch method is capable of providing
good resolution if data length samples are selected optimally. The PSD based on both
Rectangular as well as Hamming window has been designed and simulated using MATLAB. It
can be observed that the Rectangular and Hamming give better results than other windows
like Bartlett, Hanning and Blackman window.
Bartlett methods provide a method to reduce the variance of the periodograms in exchange
for a reduction of resolution in compared to standard periodograms. Blackman method is
use to correlation samples prior to transformation to reduce variance of the estimator. The
variance of the periodograms is reduced by applying a window to the autocorrelation
estimate to decrease the contribution of unreliable estimates to the periodograms. The
Welch Method used a modified version of Bartlett method in which the portion of the series
contributing to each periodograms are allowed to overlap [2]. Resolution is the ability to
discriminate spectral feature and is a key concept on the analysis of spectral estimator
performance.
I. INTRODUCTION
In this paper, we show the effect of data length on power spectral density by the help of
Welch method in rectangular and hamming window. The nonparametric Welch method in
which the power of any input is guesstimation at different frequencies [1]. It is an
improvement on the periodograms (a method of estimating the autocorrelation of finite
length of a signal) spectrum estimation method where signal to noise ratio is high and
reduces noise in the estimated power spectra in exchange for reducing the frequency
resolution. Periodogram is easy to compute and have limited ability to produce accurate
power spectrum estimation. It is biased when dealing with finite windows. As data length
increase, the rate of fluctuation in this is also increase. The rectangular window has good
resolution characteristics for sinusoids of comparable strength. The hamming window is use
to minimize the maximum nearest side lobes. The Rectangular window has admirable
resolution characteristics for sinusoids of comparable strength.
II. POWER SPECTRAL DENSITY ESTIMATION
PSD is most important application area in Digital Signal Processing. There are mainly two
types of power spectrum estimation (PSD) method: Parametric and nonparametric.
Parametric or non-classical methods an analyzed process is eplace by an appropriate model
with known spectrum. Non-parametric do not make any assumption on the data generating
process. It is start by estimating autocorrelation sequence from a given data. The power
spectrum then is estimated via FT of an estimated autocorrelation sequence. Window
function is a mathematical function that is zero valued outside of some chosen period.
When another unction or waveform or data sequence is multiplied by a window function,
the product is also zero-valued outside the period; all that is left is the part where they
overlap the observation through the window. It is simple to apply and understand. In this
method the frequency of a filter, HD(w) and the corresponding impulse response, hD(n),are
related by the inverse Fourier transform:
H0(n)= 1/2pi[H0(w)e^jwn dw ………..1
Where HD(w) is frequency response of a filter and hD(n) is corresponding impulse response.
The subscript D is used to make a difference between the ideal and practical responses.
Here, HD(w) can be obtained from hD(n) by evaluating the inverse Fourier transform. The
truncation of hD(n) to a length M-1 is equivalent to multiplying hD(n)by a rectangular
window [2] defined as
W (n) = {1, n=0, 1, 2…….M-1 (2)
And unit impulse response
H (n) =h0 (n) w (n)
H (n) = {h0 (n), n=0, 1, 2……M-1 (3)
Frequency domain function in representation of window function is,
W(w) = w(n)e^-jwn (4)
The Rectangular window has admirable resolution characteristics for sinusoids of

comparable strength. When the signal is harmonically related to the window length in first
harmonic frequency corresponding to a record length, the signal appears periodic and
infinite and is faithfully reproduced even by the rectangular window. The individuality of it
play a significant role in establishment of the resulting frequency response of the finite
impulse response filter obtained by truncation hD(n) to length M. The undesirable effects
are best alleviated by the use of window that do not contain abrupt discontinuities in their
time domain characteristics and have likewise low side lobes in their frequency domain
characteristics. For the [2] same value of M for both Rectangular and Hamming window or
other windows, the width of the main lobe is also wider for these windows compared to the
rectangular window. The Fourier transform of rectangular window:
W (w) =e^-jwn
=1-e^jwM/1-e^jw = e^-jw (M-1)/2 sin (wM/2)/sin (w/2) (6)
The Hamming window function in time domain decrease more gently towards zero on
either side and in frequency domain, the amplitude of the main lobes is wider
approximately double than that of rectangular window, but side lobes are lesser relative to
the main lobe about 40 dB down the main lobes, compared with 14 dB for the rectangular
window. Hamming window lead to a filter with wider transition width but higher stop band
attenuation.
Where Y (n) is output signal x (n) is input signal, and w(n) [1] is window function.
W (n)=00.54+0.46cos(2pin/N){-N/2<n<N/2} (N even)
=-(N-2) <n<N-1/2 (N odd)………….(8)
F=3.3/n………...(9)
Transition width, (9) Where N is filter length, and ∆f is normalized transition width Window
Length, Where n is window size. Bartlett Method in which the final estimate of the spectrum
at a given frequency is obtained by averaging the estimates from the periodograms
resultant from a non-overlapping portion of the original series. In this, the n point sequence
is subdivided in to K no overlapping segment where each segment has length L. this result in
K data segment. Power spectrum estimation for Bartlett method [3]:
In the Blackman-Tukey, we use auto-correlation and FT to get PSD in effect smooth out the
periodogram, it has better variance and better precision than Bartlett method. Power
spectrum estimation for Blackman- -Turkey method.
III. WELCH METHOD

PSE is most important application area in Digital Signal Welch method have two basic
modifications to the Bartlett method. These are allowed the data length to overlap. The
data segment can be represented as
(13) Where if is the starting point for the ith sequence. If D = M, the segment do not overlap
and the L of data sequence is identical to the data segment of Bartlett method. The second
change in Welch method is to window the data segments prior to computing the
periodogram
Where U is a normalization factor for the power
The Welch power spectrum estimate is the average of modified period gram, is
Mean value of Welch estimate
The resolution of estimated power estimation is determine by the spectral resolution of

each segment which is of length L. it is window dependent.
Program:
fs = 1000;
t = 0:1/fs:3;
x = sin(2*pi*200*t)+sin(2*pi*400*t);
figure(1)
subplot(2,1,1)
pwelch (x,[],[],[],fs)
title (‘overlay plot of 50 welch estimates’)
subplot (2,1,2)
pwelch (x,hanning(512),0,512,fs)
title(‘N=512 overlap=50% hanning’)
figure(2)
subplot(2,1,1)
pwelch(x,[],[],[],fs)
title(‘overlay plot of 50 welc estimates’)
subplot(2,1,2)
pwelch(x,hamming (512),0,512,fs)
title (‘N=512 overlap=50% hamming’)
figure (3)
subplot(2,1,1)
pwelch(x,[],[],[],fs);
subplot(2,1,2)
pwelch (x, bartlett (512),0,512,fs)
ttle(‘N=512 overlap=50% bartlett’)
figure(4)
subplot(2,1,1)
pwelch(x,[],[],[],fs);
subplot (2,1,2)
pwelch (x,[],[],[],fs);
subplot (2,1,2)
pwelch (x,boxcar (512),0,512,fs)
title (‘N=512 overlap = 50% rectangular’)

RESULT:
In this paper we are trying to show the data length effect on resolution.
For this we are considering the data sequence of 301 samples and fs=1000 in rectangular
Window and hamming window:
CONCLUSION:
In this paper, the Welch method gives the result on power spectral density with
different windows in window method. In dsp, Welch method is used to find the PSD of a
signal with reducing the effect of noise. In our paper special techniques like window
function and hamming, hanning, Bartlett, and rectangular window to extract the unwanted
noise from signal. The sampling frequency 1000 Hz fixed and the number of sample are
301.the graph shows the variation according to the different window. When data length is
short Blackman tuckey method is better than Welch method but as the data length increase
Welch method gives excellent results. The rectangular and Bartlett window have the clear
peak in the graph showing the power spectrum estimation.

Welch PSD Estimator Using Winows: SL - No. Title

Uploaded by

Copyright:

Available Formats

Welch PSD Estimator Using Winows: SL - No. Title

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Welch PSD Estimator Using Winows: SL - No. Title

Uploaded by

Copyright:

Available Formats

WELCH PSD ESTIMATOR USING WINOWS

H0(n)= 1/2pi[H0(w)e^jwn dw ………..1

W (n) = {1, n=0, 1, 2…….M-1 (2)

And unit impulse response

H (n) =h0 (n) w (n)

H (n) = {h0 (n), n=0, 1, 2……M-1 (3)

Frequency domain function in representation of window function is,

W(w) = w(n)e^-jwn (4)

The Rectangular window has admirable resolution characteristics for sinusoids of

=1-e^jwM/1-e^jw = e^-jw (M-1)/2 sin (wM/2)/sin (w/2) (6)

=-(N-2) <n<N-1/2 (N odd)………….(8)

III. WELCH METHOD

Where U is a normalization factor for the power

Mean value of Welch estimate

The resolution of estimated power estimation is determine by the spectral resolution of

title (‘overlay plot of 50 welch estimates’)

title(‘N=512 overlap=50% hanning’)

title(‘overlay plot of 50 welc estimates’)

title (‘N=512 overlap=50% hamming’)

title (‘overlay plot of 50 welch estimates’)

pwelch (x, bartlett (512),0,512,fs)

ttle(‘N=512 overlap=50% bartlett’)

title (‘overlay plot of 50 welch estimates’)

title (‘overlay plot of 50 welch estimates’)

pwelch (x,boxcar (512),0,512,fs)

title (‘N=512 overlap = 50% rectangular’)

Window and hamming window:

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.