How to separate signals with diffrent frequencies
Afficher commentaires plus anciens
t = linspace(0, 1, 1001);
x=sin(2*pi*t*3) + sin(2*pi*t*6) + sin(2*pi*t*10)+sin(2*pi*t*13)+sin(2*pi*t*20)+sin(2*pi*t*50)+sin(2*pi*t*100)+sin(2*pi*t*200);
% my signal with frequencies 3,6,10,13,20,50,100,200 KHz
sigma=0.1;
y=x+randn(size(x)).*sigma; % adding noise
I filtered the signal using fft.
I want to separate signals.
Eg : from my original signal, I want a signal with frequency range say 3 to 8 KHz, another signal say from 100 to 200 KHz. How can I separate them and display those signals.How can I do this. Anybody Please help me to solve this.?
2 commentaires
mouh nyquist
le 1 Jan 2015
you must read about chebishev and butter filter
aarthi arunachalam
le 11 Mar 2016
i too want to seperate certain frequency range from my original signal. can you help me?
Réponse acceptée
Shoaibur Rahman
le 1 Jan 2015
Add the following lines of code at the bottom of your code (after y=x+randn(size(x)).*sigma;):
ffty = fft(y);
ffty = abs(ffty(1:ceil(length(y)/2)));
ffty(ffty<100)=0;
[~,locs] = findpeaks(ffty);
freqs = (locs-1)/t(end)
signal_1 = sin(2*pi*t*freqs(1))+sin(2*pi*t*freqs(2))+sin(2*pi*t*freqs(3)); % 3-10
signal_2 = sin(2*pi*t*freqs(6))+sin(2*pi*t*freqs(7)); % 50-100
subplot(211), plot(t,signal_1)
subplot(212), plot(t,signal_2)
10 commentaires
Shoaibur Rahman
le 1 Jan 2015
ffty(ffty<100)=0; is for thresholding the data at 100.
To do this for any other data, first see the plot(abs(fft(y))), and select the threshold value so that no peak is left away.
FRom the above code, the plot(abs(fft(y)))gives the following output: BY seeing , what value can be set as a threshold. Peaks are around 500.
For my data, used plot(abs(fft(y)), got the following output
What to do?
Shoaibur Rahman
le 1 Jan 2015
Modifié(e) : Shoaibur Rahman
le 1 Jan 2015
Okay, for some reason findpeaks function does not detect the peak at 1. You can also set your threshold inside the findpeaks function. So, modify the code a little bit:
ffty = abs(fft(y));
ffty = ffty(1:ceil(length(y)/2));
ffty = fftshift(ffty);
[~,locs] = findpeaks(ffty,'threshold',100);
freqs = (locs-1)/t(end)-floor(length(ffty)/2)
For previous signal: you can use threshold at any value below the maximum in the plot, like 100, 200, 300, 400........(make sure to keep the peaks, but remove the noise).
For next signal: threshold value can be 0.2, 0.3.......
Also, for aperiodic signal like your second one, use t(end)=1
Shoaibur Rahman
le 1 Jan 2015
The data seems like a noise (may not be). Also, if you take the fft of it, you will see only one peak at 0. This means the frequency of this signal is 0. So, I think, you are getting correct result.
However, in case this is not a noise, don't leave the smaller peaks when thresholding. The fft peak locations in frequency domain are the frequencies of a signal. If you force them to be zeros by thresholding, then certainly you will miss those frequencies.
Srey
le 1 Jan 2015
Srey
le 1 Jan 2015
Shoaibur Rahman
le 1 Jan 2015
Modifié(e) : Shoaibur Rahman
le 1 Jan 2015
Use any sinusoidal function to generate a signal with that frequency, and then plot that against time. For example:
t = .....
out1 = sin(2*pi*f1*t);
out2 = cos(2*pi*f2*t);
etc....
This is just from the Fourier analysis, that states any signal (with some criteria) can be represented as summation of infinite sinusoidal functions. So, to represent only one frequency, use one sinusoidal function, or to represent multiple frequency components, use summation of multiple sinusoidal functions with those frequencies.
Srey
le 1 Jan 2015
Plus de réponses (1)
Pourya Alinezhad
le 1 Jan 2015
1-you can use several filters with different center frequencies (chebishev or butter filter)
h = fdesign.bandpass('N,F3dB1,F3dB2', N, Fc1, Fc2, Fs);
Hd = design(h, 'butter');
filter (Hd,input_signal);
2- you can use an adaptive filter and feed the signal and the desired signals to it (LMS filter).
2 commentaires
Srey
le 1 Jan 2015
Pourya Alinezhad
le 1 Jan 2015
assign Fc1, Fc2 values as 3 and 10Kz for a filter. then make another filter with 50 to 100 kHz spectrum.
Catégories
En savoir plus sur Frequency Transformations dans Centre d'aide et File Exchange
le 1 Jan 2015
le 11 Mar 2016
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
