# how to get real frequency components using MUSIC compute frequency?

9 vues (au cours des 30 derniers jours)
evelyn le 26 Nov 2023
Commenté : Chunru le 30 Nov 2023
I used pmusic ( https://au.mathworks.com/help/signal/ref/pmusic.html ). It seems that pmusic cannot get real frequency for the signal.
x = cos(0.257*pi*n) + sin(0.2*pi*n) + 0.01*randn(size(n))
should have frequency components of 0.1258 and 0.1.
1. pmusic with No Sampling Specified example
But, the peaks for x-axis(normalized frequency) are 0.257 and 0.2.
2. Specifying Sampling Frequency and Subspace Dimensions example
But, the peaks for x-axis( frequency) are 812(=0.2*fs/2) and 1258(=0.257*fs/2) .
It seems that pmusic cannot get real frequency for the signal? So the peaks at x-axis is not the real frequency? what is the actually frequency? how could I get them for pmusic
##### 4 commentairesAfficher 2 commentaires plus anciensMasquer 2 commentaires plus anciens
Walter Roberson le 26 Nov 2023
When you pass fs of 8000 into the first pmusic() call, you are lying completely about the sampling frequency.
evelyn le 26 Nov 2023
how should I set the fs for the first pmusic call?
x = cos(0.257*pi*n) + sin(0.2*pi*n) + 0.01*randn(size(n)) have frequency components of 0.1258 and 0.1.
when I set fs= 2*0.1258=0.257, the result of frequency still seems not right?
rng default
n = 0:199;
x = cos(0.257*pi*n) + sin(0.2*pi*n) + 0.01*randn(size(n));
[P,f] = pmusic(x,[Inf,1.1],[],0.257,7); % Window length = 7
plot(f,20*log10(abs(P)))
xlabel 'Frequency (Hz)', ylabel 'Power (dB)'
title 'Pseudospectrum Estimate via MUSIC', grid on

Connectez-vous pour commenter.

### Réponses (1)

Chunru le 27 Nov 2023
rng default
n = 0:199;
f1 = 0.257/2; f2 = 0.2/2;
fs= 2; % fs (fs>2*max(f1, f2)
% generate signal according to f1, f2, fs
t = n/fs;
x = cos(f1*2*pi*t) + sin(f2*2*pi*t) + 0.01*randn(size(t));
nfft=1024;
[P,f] = pmusic(x,4,nfft,fs,7); % Window length = 7
plot(f,20*log10(abs(P)))
xlabel 'Frequency (Hz)', ylabel 'Power (dB)'
title 'Pseudospectrum Estimate via MUSIC', grid on
xline([f1 f2])
##### 4 commentairesAfficher 2 commentaires plus anciensMasquer 2 commentaires plus anciens
evelyn le 30 Nov 2023
how to choose a proper fs and window length?
when I set fs to 0.4, what is the proper window length?
Chunru le 30 Nov 2023
Fs should be at least 2x of the max freq of interest to satisfy the sampling requirement.
Window length should be long enough to resolve the frequencies. For conventional method using FT to estimate spectrum, L is proportial to fs/delta_f. For the example above:
f1 = 0.257/2; f2 = 0.2/2;
fs= 3;
% Length of windows for conventional FT
fs/abs(f2-f1)
ans = 105.2632
For MUSIC, the windown length can be significantly smaller. The exact number is subject to SNR, noise properties, signal properties and so on (since signal and noise subspaces are estimated from the data).

Connectez-vous pour commenter.

### Catégories

En savoir plus sur Subspace Methods dans Help Center et File Exchange

### Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by