Sound Processing in Matlab: I

[blue_raver22]

I need help in extracting the fundamental frequency from a recorded wav file. The input is an A-chord from the guitar sampled at 44100 Hz using mono.

my source code is as follows. The variable framed is components of the signal required to be processed. So if the waveform of chord A starts at time 1.2 seconds and starts to dissipate at 3.4 seconds, framed would be the signal components from 1.2 until approximately 2.5 seconds.

y = fft(framed);
m = abs(y); % absolute value to eliminate the imaginary components
blocksize = 8000;

% Avoid taking the log of 0.
index = find(m == 0);
xfft(index) = 1e-17;

mag = 20*log(m); % take the magnitude of the fft in dB
mag = mag(1:floor(blocksize/2));
f = (0:length(mag)-1)*Fs/blocksize;
f = f(:);

% Now create a frequency vector for the x-axis and plot the magnitude and phase.
%f = (0:length(framed)-1)'*100/length(framed);
figure;plot(f,mag),
ylabel('Abs. Magnitude'), grid on
xlabel('Frequency [Hertz]')
[ymax,maxindex]= max(mag); % takes the max values
maxindex
ymax

When running this program, the desired output waveform does not show the frequency of chord A (440 Hz) but rather something else. Please help me! Our thesis doesn't seem to be progressing anymore because of this minor but irritable problem. If anyone can please help me on this. Thanks!

miguel.mcervantes@gmail.com

or

blue_raver19@yahoo.com

 

[blue_raver22]

Help!

Or an alternative method would be a simpler way in extracting the fundamental frequency from a sound file... my syntax doesn't seem to give me the desired results...

 

[lanjarote]

You can't dilucidate the fundamental frequency by selecting the peak frequency as you are doing.
Moreover when you have a chord with several fundamental frequency mixed

(you have not a self coherent signal)
but a signal composed by three fundamental (temperated and therefore
non congruent) frequencies

good luck , I am afraid that the process you wish to program is somewhat more complex.

 

[blue_raver22]

we were able to display the right frequencies already. but now its just a matter of determining the right chord after recognizing the many harmonics present in one chord... do u think i can teach the program to recognize a series of harmonics that would yield the right chord? Like the D chord has D3, A3, D4, F#4,A4, D5 and F#5... if i teach the program to learn this progression it could possibly detect the correct chord... is this a good approach or does it involve a bit more than simpy looking for a chord's harmonics?
thanks!

 

[lanjarote]

The temperated grid in which you try to locate the armonics of the notes that make up the chord are too coarse to reflect the armonc composition of the notes (due to the fact that armonics are growing lineary and temperated notes are growing logaritmically)
I think the problem you are trying to solve are quite more complex and you'd better do trying some autoscalar logaritmic transformation as wavelet o something similar.
Anyway, you could use (linear) Fourier transform but then you'd need to solve a difficult coherence problem ( that is extracting the Maximun Common Divisor of all the significant frecuencies in the signal) but this is more easy to say that to solve because if you have some noise in each frequency, you'll end up saying that the fundamental frecuency is the lower in the system; that ist the inverse of the duration of the sample.