357x Filetype PDF File size 0.77 MB Source: www.audiolabs-erlangen.de
Aspects of Music
Lecture Piece of music
Music Processing
Chord Recognition
Rhythm
Meinard Müller Melody
International Audio Laboratories Erlangen
meinard.mueller@audiolabs-erlangen.de
Harmony
Harmony: The Basis of Music Musical Chords
Pachelbel’s Canon
Combination of three or more tones which sound
simultaneously
Chord classes
– Triads including major, minor, diminished, augmented chords
– Many other more complex chords such as seventh chords
Here: focus on major and minor triads
Coversong Die Eine (Die Firma)
Musical Chords Musical Chords
The C major chord The C minor chord
Derived from the C major scale Derived from the C minor scale
C ---- the root C ---- the root
E ---- the (major)third Eb---- the (minor)third
G ---- the fifth G ---- the fifth
Musical Chords Chord Recognition
Structure of the 24 major/minor chords
Development of automatic methods for the
harmonic analysis of audio data
Applications in the field of music information
01234567891011 retrieval:
CC#DD#EFF#GG#AA#B – music segmentation
C major – cover song identification
– audio matching
C minor – music structure analysis
– …
Chord Recognition Chord Recognition
Signal Result Signal Feature Classification Result
Extraction
Segmentation Chroma features Pattern matching
Chord label assignment C F G C Type Template
Resolution Gaussian
Compression Hidden Markov Models
Smoothing Graphical Models
Chord Recognition Baseline Method for Chord Recognition
Given: Audio file Chord templates 24 major/minor chords
Output: Segmentation and chord labeling C C# D D# …… C C# ……
major major major major minor minor
B 0 0 0 0 …… 0 0 ……
A# 0 0 0 1 …… 0 0 ……
A 0 0 1 0 …… 0 0 ……
G# 0 1 0 0 …… 0 1 ……
G 1 0 0 1 …… 1 0 ……
F# 0 0 1 0 …… 0 0 ……
C A:min E:min FC FCA:min E:min
G G G F 0 1 0 0 …… 0 0 ……
E 1 0 0 0 …… 0 1 ……
D# 0 0 0 1 …… 1 0 ……
D 0 0 1 0 …… 0 0 ……
CGA:minE:min G C# 0 1 0 0 …… 0 1 ……
F C FG F CF
C 1 0 0 0 …… 1 0 ……
Baseline Method for Chord Recognition Baseline Method for Chord Recognition
Chord templates 24 major/minor chords 24 chord templates
Chromafeature
(12 major, 12 minor) extraction (framewise)
Chroma
Chord labels
Baseline Method for Chord Recognition Baseline Method for Chord Recognition
24 chord templates Chromafeature 24 chord templates Chromafeature
(12 major, 12 minor) extraction (framewise) (12 major, 12 minor) extraction (framewise)
Compute for each frame the distance of Compute for each frame the distance of
the feature vector to the 24 templates the feature vector to the 24 templates
Selected chord according to template with
minimal distance to respective feature vector
Problems in Chord Recognition Problems in Chord Recognition
Problem: Transitions between subsequent chord Problem: Monphonic musical passages
Example: Chopin Mazurka Op. 68 No.3 Example: Excerpt of Wagner’s Meistersinger
Chromagram
Correct 1
B
False positive A# 0.9
A 0.8
False negative G# 0.7
G 0.6
F#
0.5
F
E 0.4
D# 0.3
D 0.2
C# 0.1
C
0
0 10 20 30 40 50 60
Problems in Chord Recognition Problems in Chord Recognition
Problem: Frame-wise chord analysis may not be meaningful Problem: Ambiguity of chords
Example: Bach: Prelude C major, BWV 846
A minor C major E minor
A
C C
E E E
Problem: Broken chords G G
B
→ Measure-wise chord analysis necessary
Problems in Chord Recognition Problems in Chord Recognition
Problem: Reduction to the 24 major/minor chords makes Problem: Tuning problems
the recognition of more complex chords difficult/impossible! Example: Excerpt of Wagner’s Meistersinger
Example: Prelude C major, BWV 846, mm.19-25
Chromagram (from MIDI)
1
B
0.9
A#
A 0.8
G# 0.7
G 0.6
F#
0.5
F
0.4
E
0.3
D#
D 0.2
C# 0.1
C 0
0 10 20 30 40 50 60
Problems in Chord Recognition Problems in Chord Recognition
Problem: Tuning problems Problem: Tuning problems
Example: Excerpt of Wagner’s Meistersinger Example: Excerpt of Wagner’s Meistersinger
Chromagram (from MIDI) Problem: Chromagram (from MIDI) Problem:
1 Audio is tuned more than 1 Audio is tuned more than
B B
A# 0.9 half a semi-tone upwards 0.9 half a semi-tone upwards
A#
A 0.8 A 0.8
G# 0.7 G# 0.7 Solution:
G 0.6 G
0.6
F# F# Adjust frequency binning
0.5 0.5
F F
E 0.4 0.4 when computing pitch
E
D# 0.3 D# 0.3 features.
D 0.2 D 0.2
C# 0.1 C#
0.1
C C
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
no reviews yet
Please Login to review.