Zero frequency in audio signal

Hello sir,

I am working on a project where one of the task is to find the FFT of an audio file that I have. The audio file basically contain sound from a 'trumpet'. It is of 5s duration.Firstly I use 'audioread' function to read the file which gave me a vector i.e a discrete time samples of audio. Then when I took fft to see the frequency of sound clip, it showed me max peak at 0 Hz. How is it possible? I am not getting it. The code works fine for other sounds.

I will share the code that I am using please tell me where I am going wrong.

[c,fs]=audioread('T4.wav');

c=c';

dft=fft(c);

dft=dft(1:length(c)/2+1);

fval=fs*(0:length(c)/2)/length(c);

[Max,I]=max(abs(dft));

plot(fval,abs(dft))

fprintf('Maximum occurs at %3.2f Hz.\n',fval(I))

Answer 1

[Star Strider]

The ‘0 Hz’ peak is the constant (or d-c, direct current) offset. To eliminate it and see the other frequencies in your signal, subtract the mean from your signal before taking the fft.

Example:

[c,fs]=audioread('T4.wav');

c = c-mean(c);

c=c';

dft=fft(c)

.. REST OF YOUR CODE ...

[Nachiket Patki]

Thank you sir. I got it.

[Star Strider]

My pleasure.

[Nachiket Patki]

Ok so now I got the desired frequencies, now there is one thing that I am looking for is to classify.I have total of 10 frequencies. I want to classify them as say Type A and Type B. Can we classify them using power spectral density or may be using STFT? or any other technique should be used in this scenario?

[Star Strider]

You have to do the classification using a classifier function, for example the fitcdiscr (link) (or classify) function to do a linear discriminant classification. There are several other classifiers. The classifier you choose depends on the problem you want to solve.

The Fourier transform and similar functions do not do classification.

[Nachiket Patki]

Ok let me try this out. I will ask doubt if I get stuck. Thank you sir.

[Star Strider]

My pleasure.