Finding max value and its index

So, here is the problem.

I have a wav file (attached), one tuts piano sound, and I analyzed it. Using this.

[wave,fs] = audioread('file01.wav');
n=length(wave)-1;

figure

t=0:1/fs:n/fs;
subplot(3,1,1), plot(t,wave)
subplot(3,1,1), xlabel('Time (Second)')
subplot(3,1,1), ylabel('Amplitude')

f=0:fs/n:fs;
wavefft=abs(fft(wave));
subplot(3,1,2), plot(f,wavefft)
subplot(3,1,2), xlabel('Frekuency (Hz)')
subplot(3,1,2), ylabel('Magnitude')

[max_value, index] = max(wavefft(:));

subplot(3,1,3), plot(f,wavefft)
subplot(3,1,3), axis([0 index 0 max_value])

The output is like this.

When I search max value using

[max_value, index] = max(wavefft(:));

I got this. Surely the peak would be on the right side, right?

When I inspect it, the max_value is in index 1111 Hz.

But when I zoom in manually, it's on index 740 Hz.

Did I do something wrong?

[Star Strider]

Seems you did everything correctly. The data in ‘subplot(3,1,3)’ is correct.

Explore the documentation on fft to understand why. (See the documentation on fftshift to understand the reason subplot(3,1,2) is misleading you.)

Answer 1

[Image Analyst]

Index 1111 is not at 1111 Hz necessarily. From the plot above, it looks like it's at about 750 Hz. But you msight want to use fftshift to shift the ends to the middle.

[Edo Christianto]

Thanks for the answer. I think I got the gist of it. Shift the ends to the middle. Incorporating fftshift to my script, from

wavefft=abs(fft(wave));

to

wavefft=fftshift(abs((fft(wave))));

what do I get is

Then, how can I know that this particular piano tuts is 750Hz based from this fftshift?

[Star Strider]

You actually don’t need to. You did everything correctly in subplot(3,1,3) in your original Question.

Use those data and that plot.

[Image Analyst]

If you shift the spectrum to put the 0 frequency at the center, you need to subtract ~2.25 * 10^4 (actually x(floor(length(x)/2)) I believe) from your x axis so that 0 shows up at the middle, with the middle index of the array you're plotting.

[Edo Christianto]

Ahh... Actually I'm so confused right now... If you don't mind, I would ask again about this in the near future.

Anyway, thanks for all the help, kind Sirs/Mesdames.

(Note for future reference.)

Answer 2

[Malcolm Hawksford]

Lets assume you have a stereo .wav file and simply want to find the peak amplitude and sample index (time domain) in each channel. Assume the file is called Cohen.wav then this works for me: [p q]=max(abs(audioread('Cohen.wav'))); p gives the peak values and q the corresponding indices.