Pitch Detection: Music, Physics, and the Brain Tom Goodman - - PowerPoint PPT Presentation

▶
pitch detection music physics and the brain
SMART_READER_LITE
LIVE PREVIEW

Pitch Detection: Music, Physics, and the Brain Tom Goodman - - PowerPoint PPT Presentation

Dec 17, 2022 1.31k likes •1.72k views

Pitch Detection: Music, Physics, and the Brain Tom Goodman University of Birmingham Wednesday 30 th January 2018 Wednesday 30 th January 2018 Tom Goodman (University of Birmingham) Research Skills 1 / 37 Contents Aim of Pitch Detection

slide-1
SLIDE 1

Pitch Detection: Music, Physics, and the Brain

Tom Goodman

University of Birmingham

Wednesday 30th January 2018

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 1 / 37

slide-2
SLIDE 2

Contents

Aim of Pitch Detection Historical Background A Little Music Theory Physics of Sound Proposed Real-Time Algorithm Future Considerations Biological Inspiration for the Future Fin.

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 2 / 37

slide-3
SLIDE 3

Background

Aim

Take in an acoustic musical signal, and output the notes present.

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 3 / 37

slide-4
SLIDE 4

Background

Aim

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 4 / 37

slide-5
SLIDE 5

Background

A Little History

history Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 5 / 37

slide-6
SLIDE 6

Background

Pythagoras & Early Music (c. 550BC)

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 6 / 37

slide-7
SLIDE 7

Background

Helmholtz & the Rise of Psychoacoustics (1800s)

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 7 / 37

slide-8
SLIDE 8

Background

Some Basic Music Theory

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 8 / 37

slide-9
SLIDE 9

Background

Categorising Notes

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 9 / 37

slide-10
SLIDE 10

Background

Categorising Notes

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 10 / 37

slide-11
SLIDE 11

Background

Musical Notes

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 11 / 37

slide-12
SLIDE 12

Background

Standing Waves

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 12 / 37

slide-13
SLIDE 13

Background

Harmonics

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 13 / 37

slide-14
SLIDE 14

Background

Implications

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 14 / 37

slide-15
SLIDE 15

Background

Timbre

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 15 / 37

slide-16
SLIDE 16

Background

The Fourier Transform

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 16 / 37

slide-17
SLIDE 17

Real-Time Algorithm

Overview

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 17 / 37

slide-18
SLIDE 18

Real-Time Algorithm

Assumption - No Undertones

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 18 / 37

slide-19
SLIDE 19

Real-Time Algorithm

Clustering

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 19 / 37

slide-20
SLIDE 20

Real-Time Algorithm

“Band-Pass” Filter

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 20 / 37

slide-21
SLIDE 21

Real-Time Algorithm

Selection Loop & α-cutoff

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 21 / 37

slide-22
SLIDE 22

Real-Time Algorithm

Notes Reduce

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 22 / 37

slide-23
SLIDE 23

Real-Time Algorithm

Testing

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 23 / 37

slide-24
SLIDE 24

Real-Time Algorithm

Results

Precision Recall Specificity Accuracy F-Score Test 1 78.09% 93.89% 83.00% 86.89% 84.78% Test 2 100.00% 97.22% 100.00% 97.85% 98.55% Test 3 82.46% 75.67% 77.52% 76.36% 78.85% Test 4 86.31% 82.50% 53.33% 75.83% 83.47% Test 5 65.08% 100.00% 64.29% 79.07% 78.47% µ 82.39% 89.86% 75.63% 83.20% 84.82%

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 24 / 37

slide-25
SLIDE 25

Real-Time Algorithm

Further Considerations

Adapt evaluation to use a larger & more widely-adopted dataset (eg. MAPS) Adapt ‘Notes Reduce’ to model harmonics on an instrument-to-instrument basis Apply source separation to reduce polyphonic problems to multiple monophonic problems Take temporal aspects of sound into account Take inspiration from biological systems for pitch detection in humans (such as the ear/brain)

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 25 / 37

slide-26
SLIDE 26

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 26 / 37

slide-27
SLIDE 27

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 27 / 37

slide-28
SLIDE 28

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 28 / 37

slide-29
SLIDE 29

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 29 / 37

slide-30
SLIDE 30

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 30 / 37

slide-31
SLIDE 31

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 31 / 37

slide-32
SLIDE 32

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 32 / 37

slide-33
SLIDE 33

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 33 / 37

slide-34
SLIDE 34

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 34 / 37

slide-35
SLIDE 35

Humans & Pitch Detection - A perfect solution?

A Human Solution

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 35 / 37

slide-36
SLIDE 36

Humans & Pitch Detection - A perfect solution?

Some Thoughts for the Future

Do the ear and brain process monophonic and polyphonic signals in different ways? How does the brain subconsciously abstract timbre? Can we build models and solutions in similar ways?

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 36 / 37

slide-37
SLIDE 37

Humans & Pitch Detection - A perfect solution?

Questions?

Tom Goodman t.a.goodman@cs.bham.ac.uk http://tomg.io/ Office 218 School of Computer Science University of Birmingham Slides at http://tomg.io/research-skills-slides.pdf Paper at http://tomg/io/isspit-1.pdf Feel free to ask any questions!

Tom Goodman (University of Birmingham) Research Skills Wednesday 30th January 2018 37 / 37