Rating Dialog
Control of pitch range and rhythm pattern; motif extraction

A view on the global structure of the fugue, the temporal distribution of the motifs for each voice. The voices congregated to trace the bandwidth of the sound.

# Name of rating Computation Rewards/left/maximum Punishes/right/minimum
0 ZeroOrder conceptually same as FirstOrder
1 FirstOrder 1st order joints that occur frequently in the database generally not a good idea
2 Tracking commonality of sequence of 4 pitches of consecutive notes in a voice sequences of 4 consecutive notes that occur frequently in the database
3 VoiceCoverage sum of all non-rests silence
4 Actions new notes, notes changing into rest, jumps +1 finger movement long rests, holding of notes
5 Interlock counts events that feature hits as well as holds scores where notes are introduced while other notes in other voices are held
6 Pitchlevel average of pitch high pitch average low pitch average
7 Envelope voices with small maximum range soprano is far apart from bass
8 Pitchrange distance of pitches of notes of each voice from the user defined preferred interval pitches are all in preferred ritch range pitches are outside these intervals
9 Together sum of symmetric difference between successive joints note sequences with small pitch changes large jumps in melodies
10 MajorMinor sum of major chords minus sum of minor chords major, happy minor, sad
11 Uniqueness -1 for joints that already occur in past score unique progressions repetitions of harmonic transitions
12 BackwardsSymmetry identical jumps +2, identical links +1 one measure and/or half a measure ago melodic and ryhmic resemblance with events in the recent past melodic and ryhmic innovation
13 Library counts ticks of colored notes high temporal coverage of notes from library material that does not resemble material from the library

Correlation of ratings ***

bachwelltemperedclavier by Philip Goeth

Rating control in the valence-arousal 2-dimensional space

mapping of emotions by James Russell

There are 10 mostly independent rating. This means 10 different ways on how to control the musical selection. In practice we are looking for a way to reduce the dimensionality. Our solution uses two publications on the correlation of emotional state to musical feature. The emotional state is characterized by a location in the valence and arousal coordinate system (by James Russell).

In his Master's thesis EEG emotion recognition and affective music BCI, Stefan Ehrlich uses the following terminology and mapping

Dimension Minimum Maximum Category
valence misery pleasure
Locrian Lydian harmonic mode
low pitch high pitch voicing
arousal sleepiness arousal
few, identifiable notes many notes, not identifiable rhythmic roughness
slow fast tempo
quiet loud relative loudness

To sweep through the harmonic modes, S. Ehrlich uses Mark Schmuckler's order:

Mode 1 2 3 4 5 6 7
Title Lydian Ionian Mixolydian Dorian Aeolian Phrygian Locrian
Cadence F B C F C F G C G C D G D G A D A D E A E A B E B E F B

The cadence corresponds to the standard I VI V I chord sequence, where

In their publication Relationships between musical structure and psychophysiological measures of emotion from 2007, Patrick Gomez and Brigitta Danuser derive a non-linear correlation between valence/arousal coordinate and musical feature through experimentation. The paper displays the variation of emotional state for the following categories: