Signal Processing: Musical Instrument Recognition & Resources

[inadaze]

Hi,
I was just wondering if anyone has any experience with signal processing. I am working with musical instrument recognition and would like some help. What is the main difference between musical instrument signals? Where can I get some resources on signal processing(I am using Matlab, but anything would be a great help).

Thanks
JAy

 

[chroot]

I do quite a lot of digital signal processing.

The reason different instruments sound different (even while playing the same note) is because each instrument produces its own blend of harmonics. The amount of power in the harmonics is a "fingerprint" of the instrument.

- Warren

 

[inadaze]

Wow, your helping me out everywhere on this forum. ha. thank you.
So, what your saying is that if I were to run an FFT on any sound and count the amount of Harmonics I would be able to tell what instrument it is?
What about a guitar for instance...depending on where the string is struck it seems that I get different sets of harmonics. Is that it, or do the same harmonics show up in the same position with only the peek height changing?

Thanks
Jay

 

[chroot]

Theoretically, there's an infinite number of harmonics from every instrument. If you strike middle C (440 Hz), you'll get a harmonic at 880 Hz, 1320 Hz, etc.

The harmonics for middle C are always in the same place, regardless of the instrument. The relative heights of the peaks is what makes a trumpet sound like a trumpet and a piano sound like a piano. Some instruments have more second harmonic, while some have more third, etc.

You couldn't detect an infinite number of harmonics in practice, of course, because eventually the very weak high harmonics will not be discernable from random noise.

- Warren

 

[inadaze]

okay, that makes sense. Now when I analyze some instruments, I get peaks at dissonant intervals (not at steady intervals of the fundamental). What are these and how could I interpret them?

Jay

 

[Njorl]

Most musical instruments resemble one of two wave problems - a string bound at both ends, or a tube open at one or both ends. However, they are not perfect. There are resonators built in, and the mouth acts as a resonator as well. If a resonator can capture one frequency, imperfections might translate that energy to another frequency.

Horns, for instance, might capture some energy from longitudinal sine waves and convert them to bessel functions at their flared ends.

I bet flutes and picolos are not producing much in the way of these anharmonic signals.

Njorl