Characterization of Different Musical Instruments using Fast Fourier Transform

1
Characterization of Different Musical Instruments
using Fast Fourier Transform

• David Hoffert
• Mo Chan
• Patrick Johanning
• ME 368 Final Project
• Spring 2007

2
Outline

• Theoretical Background
• Objective
• Equipment
• Procedure
• Data
• Analysis and Results
• Summary

3
Theoretical Background

• Each note specific fundamental frequency
• Different instruments sound different, even when
  playing the same note
• What makes this difference?

4
Theoretical Background

• Harmonic frequencies complement fundamental
  frequency
• These vary in magnitude for each instrument
• Also have overall frequency trends

5
Objective

• Characterize different musical instruments based
  on their frequency spectra
• Measure frequency spectra for instruments in
  brass, woodwind, strings, and voice classes
• Plot trends in harmonic frequencies and overall
  frequency data

6
Equipment

• Same as for the acoustics lab
• Microphone ? DAQ ? LabView FFT
• Low uncertainty experiment!

7
Procedure

• Each instrument plays middle C scale 3x
• Record background noise between each
• 40000 Hz, 20000 samples 2 Hz resolution
• Subtract out background noise, normalize to first
  peak to eliminate volume concerns
• File size 40 MB each!

8
Subtracting out background noise
It does make a difference!
Construction noise is annoying!
9
Data
10
Normalization

• Clarinet first peak 50 dB, second 60 dB
• Trumpet first peak 80 dB, second 85 dB
• Wrong trumpet higher than clarinet
• Overall sound level must be accounted for
• Clarinet 2nd 120, Trumpet 2nd 106
• Clarinet is higher than trumpet!

11
Analysis and Results

• Manually identify first seven peaks
• Seven because guitar stops after that
• Others have more peaks, impacts sound

12
Analysis and Results

• Can also reconstruct sound wave

13
Summary

• Same note, same fundamental frequency
• Harmonics and overall trends make the difference
• Can see obvious differences with just one
  instrument from each class