music analysis
DESCRIPTION
Josiah Boning TJHSST Senior Research Project Computer Systems Lab, 2007-2008. Music Analysis. Purpose. Framework for study of audio data Components: Statistical Signal Processing Machine Learning Uses: Aircraft identification Speech recognition and synthesis And more... - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/1.jpg)
Music Analysis
Josiah Boning
TJHSST Senior Research ProjectComputer Systems Lab, 2007-2008
![Page 2: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/2.jpg)
Purpose Framework for study of audio data Components:
Statistical Signal Processing Machine Learning
Uses: Aircraft identification Speech recognition and synthesis And more...
Ideal: computer recognition of music
![Page 3: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/3.jpg)
Background
Bigarelle and Iost (1999) Music genre can be identified by fractal dimension
Basilie et al. (2004) Music genre can be identified by machine learning
algorithms Used discrete MIDI data
![Page 4: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/4.jpg)
Methods
Data Processing Spectral Analysis: Fourier Transform Fractal Dimension: Variation and ANAM Methods
Machine Learning Feed-Forward Neural Network
![Page 5: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/5.jpg)
Spectral Analysis – Fourier Transform
F v =∫−∞
∞f t e−2 pi i v t dt
Time domain to frequency domain
![Page 6: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/6.jpg)
Fourier Transform
![Page 7: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/7.jpg)
Spectrogram
Many Fourier transforms sequentially
![Page 8: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/8.jpg)
Frequency Aggregate
Horizontal sum of spectrogram
![Page 9: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/9.jpg)
Magnitude Aggregate
Vertical sum of spectrogram
Can perform second Fourier transform
![Page 10: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/10.jpg)
Fractal Dimension
lim0
2−
log 1b−a∫a
b
∣max f t ∣x−t∣−min f t
∣x−t∣ ∣dx log
lim0
2−
log 1b−a ∫
x=a
x=b [ 1
2 ∫t 1=0
∫t 2=0
∣ f xt1− f x−t 2∣dt1dt2 ]
1 /
dx log
Bigerelle and Iost – used to distinguish genre Variation Method:
ANAM Method:
![Page 11: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/11.jpg)
Fractal Dimension
Inaccurate calculations Correct values around 1.6-1.9 Variation: ~1.16 ANAM: ~2.25
Interpolation between samples didn’t help
![Page 12: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/12.jpg)
Machine Learning
Neural networks Feed-Forward
![Page 13: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/13.jpg)
Neurons
Each node performs weighted, rescaled sum of input values
Scaling: Sigmoid function
f x= 11e−x
![Page 14: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/14.jpg)
Results
Frequency aggregate – songs are similar!
![Page 15: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/15.jpg)
Results
Magnitude and frequency: negative correlation Not shared by
white noise
Fourier transform of magnitude aggregate Meaningless
![Page 16: Music Analysis](https://reader034.vdocuments.net/reader034/viewer/2022051316/56815715550346895dc4b3b8/html5/thumbnails/16.jpg)
Where next?
Move code from driver to Fourier module Otherwise, well organized
Fix fractal dimension calculations Neural network learning algorithms Beat detection Other signal processing techniques