SLIDE 1
MIR FOR CLASSICAL MUSIC Dasaem Jeong PhD Student PhD student in - - PowerPoint PPT Presentation
MIR FOR CLASSICAL MUSIC Dasaem Jeong PhD Student PhD student in - - PowerPoint PPT Presentation
MIR FOR CLASSICAL MUSIC Dasaem Jeong PhD Student PhD student in MAC Lab Bachelor in Mechanical Engineering Master in Culture Technology ... in KAIST What do I research? Classical Music Delivering to listeners Genre-specific
SLIDE 2
SLIDE 3
What do I research?
SLIDE 4
Classical Music Delivering to listeners
SLIDE 5
Genre-specific Research Each genre has its own characteristics It helps to make more assumption
SLIDE 6
SLIDE 7
SLIDE 8
SLIDE 9
Characteristics of classical music How can we put these two music in one category?
SLIDE 10
The Score
SLIDE 11
Even 4’ 33’’ has score
SLIDE 12
There is no authentic performance
- r original recording of the piece
SLIDE 13
Score (sheet) Variety in performance
SLIDE 14
Audio-score alignment
SLIDE 15
Audio-to-Score Alignment
Score Score Audio Target Audio
SLIDE 16
Audio-to-Score Alignment
Score Score Audio Target Audio
SLIDE 17
Audio-to-Score Alignment Calculate similarity by comparing every frame-by-frame Dynamic Time Warping
SLIDE 18
Automatic Accompaniment
SLIDE 19
Performance Analysis
SLIDE 20
What is the secret of great pianists?
Whitestone Photo
SLIDE 21
Vladimir Horowitz
SLIDE 22
Tempo Intensity
SLIDE 23
Limitation of previous research Intensity is calculated in beat-level Assumption: Simultaneously played notes have same intensity
SLIDE 24
Estimating intensity of individual note
Basis Activation Spectrum Basis
≈
X V ≈ WH
Spectrum Basis: Spectrum of each piano key Basis Activation: Activation of each piano key over time
SLIDE 25
Result Melody shows more intensity than accompaniment part.
SLIDE 26
Application
SLIDE 27
PerformScore
SLIDE 28
MIDI XML
SLIDE 29