List of Feature Definition Files¶
Features are organized and bundled into Feature Definition Files (FDF). Each of the 92 FDFs contains several single features which can be selected or de-selected individually. Moreover, FDFs are sorted in different categories with respect to the character and nature of the contained features. Accent features are mostly equivalent to binary structural markers for positions on which tonal events might be perceived as more salient than others. Auxiliary features are no features in the strict sense, but are useful for advanced and combined computations. Tone Formation features deal with articulations (staccato, legato etc.) and annotated modulations . Contour features try to capture aspects of a melody contour in time vs. pitch space. Interval features derive from differences of pitch and pitch class values. They come in four different levels of exactness: semitone intervals, fuzzy intervals (aka ‘refined contour’), Parson’s Codeyear (aka ‘contour’), as well as interval classes. Metadata features access annotated metadata of melodies. Meter features deal with metrical annotations and derived transformations. Rhythm deals with the time componente of melodies mostly durations and inter-onset intervals. Sequence features come in three different flavours: interval, pitch and rhythm. They try to catch aspects of sequentiality, such as bi- and trigrams and runlengths. Finally, Structure features are based on self-similarities of melodies on the level of individual phrases.
Note
FDFs are mostly of two different varieties: Those bundling single, scalar-valued features, and those containing vector or matrix-valued data. Vector-valued features are often either histograms or event-based features, which return for each event a certain value, e.g., a list of raw values of a certain transformation. For a better understanding of the transformation underlying many features, a look into basic transformations might be helpful.
Accents¶
Feature Bundle |
---|
Auxiliary¶
Feature Bundle |
---|
Contour¶
Feature Bundle |
---|
Interval¶
Feature Bundle |
---|
MLA¶
Feature Bundle |
---|
Metadata¶
Feature Bundle |
---|
Meter¶
Feature Bundle |
---|
Pitch¶
Feature Bundle |
---|
Rhythm¶
Feature Bundle |
---|
Sequence/Interval¶
Feature Bundle |
---|
Sequence/Pitch¶
Feature Bundle |
---|
Sequence/Rhythm¶
Feature Bundle |
---|
Structure¶
Feature Bundle |
---|
Tone Formation¶
Feature Bundle |
---|
Accents¶
Interval jump accents¶
Description: Large pitch jumps are thought to convey accents, i.e., have a higher salience. However, no consensus is reached what means ‘large’ nor if it should be absolute or relative jumps nor if the tone before or after the jump or both should receive the accent. Some basic variants are collected here, but further research is needed to settle the issue.
Feature Definition File Name:
ACCENTS_JUMPS
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0,1] |
N |
Raw vector of accents for tones following a large pitch jump of at least 3 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones following a large pitch jump of at least 4 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones following a large pitch jump of at least 5 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones before a large pitch jump of at least 3 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones before a large pitch jump of at least 4 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones before a large pitch jump of at least 5 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones before and after a large pitch jump of at least 3 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones before and after a large pitch jump of at least 4 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones before and after a large pitch jump of at least 5 semi-tones (either direction). |
|
Integer [0,1] |
N |
Raw vector of accents for tones after a pitch interval that is at least 1 semi-tone larger than the previous interval. |
|
Integer [0,1] |
N |
Raw vector of accents for tones after a pitch interval that is at least 2 semi-tone larger than the previous interval. |
|
Real [0, 1] |
N |
Raw vector of accents according to Thomassen’s algorithm (1982), which is based on the seven possible pitch direction patterns that can be formed by 2-interval chains (3-note patterns). Values are not binary, but probabilities. |
|
Inetger[0,1] |
N |
Raw vector of Thomassen accents with exceed a certain threshold (.75 in this case). |
Duration accents (Abs. IOI)¶
Description: Durations accent are based on the music-psychological observation, that for short-long (SL) patterns the second tone is perceived as accented (more pronounced, more salient). Included here are several accent features that operationalize the intuitive notion of short-long in several different ways based on IOI classification according to a absolute reference time of 0.5 s.
Feature Definition File Name:
ACCENTS_DURATION_ABS
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0,1] |
N |
Raw accent vector of tones with an higher (longer) IOI class than the previous one. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class at least two classes away from that of the previous one. |
|
Integer [0,1] |
N |
Raw accent vector of tones with a higher/longer IOI class than the most frequent (mode) IOI class. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class at least two classes away from the most frequent (mode) IOI class. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class longer than mode IOI class of the past 5 IOIs. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class at least two classes longer than the mode IOI class of the past 5 IOIs. |
Structural accents (phrases)¶
Description: Structurally important positions can give rise to accents (and vice versa). Currently, only phrases markers are implemented here.
Feature Definition File Name:
ACCENTS_STRUCTURE
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0,1] |
N |
Raw vector of markers for phrase beginnings (if available). |
|
Integer [0,1] |
N |
Raw vector of markers for phrase endings (if available). |
|
Integer [0,1] |
N |
Raw vector of markers for phrase beginning & endings (if available). |
Duration accents (Rel. IOI)¶
Description: Durations accent are based on the music-psychological observation, that for short-long (SL) patterns the second tone is perceived as accented (more pronounced, more salient). Included here are several accent features that operationalize the intuitive notion of short-long in several different ways, but always based on IOI classification according to a relative reference time of beat duration.
Feature Definition File Name:
ACCENTS_DURATION_REL
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0,1] |
N |
Raw accent vector of tones with an higher (longer) IOI class than the previous one. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class at least two classes away from that of the previous one. |
|
Integer [0,1] |
N |
Raw accent vector of tones with a higher/longer IOI class than the most frequent (mode) IOI class. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class at least two classes away from the most frequent (mode) IOI class. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class longer than mode IOI class of the past 5 IOIs. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an IOI class at least two classes longer than the mode IOI class of the past 5 IOIs. |
Pitch extrema accents¶
Description: Extreme pitch values (maxima and minima) are likely positions to stand out from the context. Some variations in operationalisation are possible (e.g., with respect to cambiatas and trillers).
Feature Definition File Name:
ACCENTS_PITCH_EXTREMA
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0,1] |
N |
Raw accent vector of pitch extrema (no restrictions). |
|
Real [0,1] |
Var |
Raw accent vector of pitch extrema (excluding proper cambiata). |
|
Real [0,1] |
N |
Raw accent vector of pitch extrema (sensu Steinbeck, with at least two intervals before and after the extrema leading strictly to it). |
Metrical accents (on/offbeat)¶
Description: Metrical accents are one the most fundamental accents. Classically, meter is partly defined as an underlying regular accent structure. Syncopations are likely to give rise to accents as well, since they violate metrical expectations. Here, on-beat as well as off-beat (i.e., syncopation) accents collected.
Feature Definition File Name:
ACCENTS_METER
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0,1] |
N |
Raw vector of accents on the primary accent (first beat) of a bar. |
|
Integer [0,1] |
N |
Raw vector of accents on the secondary accent of a bar, if present, e.g. in on the 3rd beat in 4/4-measures (hence the name). |
|
Integer [0,1] |
N |
Raw vector of accents on the primary and secondiry accent of a bar. |
|
Integer [0,1] |
N |
Raw vector of accents on all beat positions in a bar. |
|
Integer [0,1] |
N |
Raw vector of accents on all syncopations right before the primary accent in a bar (‘anticipated 1’). |
|
Integer [0,1] |
N |
Raw vector of accents on all syncopations right before the secondary accent in a bar (‘anticipated 3’). |
|
Integer [0,1] |
N |
Raw vector of accents on all syncopations right before primary and secondary accents in a bar (‘anticipated 1s and 3s’). |
|
Integer [0,1] |
N |
Raw vector of accents on all syncopations right before all beats in a bar. |
|
Integer [0,1] |
N |
Raw vector of accents on all syncopations on every sub-beat metrical level (i.e., excluding half-beat level). |
|
Integer [0,1] |
N |
Raw vector of structural markers for swing candidates. |
Duration accents (Raw IOI)¶
Description: Durations accent are based on the music-psychological observation, that for short-long (SL) patterns the second tone is perceived as accented (more pronounced, more salient). Icnluded here are several accent features that operationalize the intuitive notion of short-long in several differnet ways based on raw IOIs.
Feature Definition File Name:
ACCENTS_DURATION
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0,1] |
N |
Raw accent vector of tones with longer IOI than the previous tone. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an at least twice longer IOI than the previous tone. |
|
Integer [0,1] |
N |
Raw accent vector of tones with a longer IOI than mode IOI of all tones. |
|
Integer [0,1] |
N |
Raw accent vector of tones with an at least twice longer IOI than mode IOI of all tones. |
|
Integer [0,1] |
N |
Raw accent vector of tones with IOI that is at least 41% longer than mean of the past 5 IOIs. |
|
Integer [0,1] |
N |
Raw accent vector of tones with IOI that is at least twice as long than mean of the past 5 IOIs. |
Harmony accents¶
Description: Structural marker for in-chord or out-of-chord tones (only for melodies with annotated chords).
Feature Definition File Name:
ACCENTS_HARMONY
Category:
Accents
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0,1] |
Var |
Raw markers of chord tones (excluding upper structures). |
|
Integer [0,1] |
Var |
Raw markers of chord tones (including upper structures). |
|
Integer [0,1] |
Var |
Raw marker vector of non-chord tones. |
Auxiliary¶
Melody Export¶
Description: Exports melody as lists of onset, duration, pitch, and metrical positions.
Feature Definition File Name:
GENERAL_MELODY_RAW
Category:
Auxiliary
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
N |
Onsets (in seconds) of events. |
|
Real |
N |
Normalized onsets [0-1] of events. |
|
Real |
1 |
First onset (in seconds) of events. |
|
Real |
1 |
Last offset (in seconds) of events. |
|
Real |
N |
Duration (in seconds) of events. |
|
Real |
N |
Raw MIDI pitch values of events. |
|
String |
N |
Metrical positions of events (MP-dot-notation). |
|
String |
Var |
ID of containing phrase. |
|
Real |
N |
Offsets (in seconds) of events. |
Sections IDs and Labels¶
Description: Returns sequences of raw phrase IDs, form labels, chorus IDs, chord labels and chord types.
Feature Definition File Name:
GENERAL_SECTION_IDS
Category:
Auxiliary
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer |
Sequence of event-wise phrase IDs. |
|
|
Integer |
Var |
Sequence of event-wise chorus IDs. |
|
Integer |
Var |
Sequence of event-wise form labels (as annotated). |
|
Integer |
Var |
Sequence of event-wise chord symbols (as annotated and normalized). |
|
Integer |
Var |
Sequence of event-wise chord type labels (triad type + seventh). |
|
Integer |
Var |
Type labels for chord sections (triad type + seventh). |
Contour¶
Waviness/Zig-Zagity.¶
Description: This feature set contains features related to (pitch-)extremal points in a melody. A local extremum (or turning) point can be either a minimum or maximum. Hence, pitches right before and right after an extremal point are both lower or equal (maximum) or higher or equal (minimum). For this module only true extremal points are counted, where the pitches must be strictly higher or lower to both sides. The percentage of extremal points can be regarded as a measure for zig-zagity (or waviness) of a melody. The extreme case are two alternatine pitches, where each and every point is a local extremum resulting in an extrema ratio of 1.
Feature Definition File Name:
PITCH_WAVINESS
Category:
Contour
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Percentage of extremal points. |
|
Integer |
1 |
Number of extremal points. |
|
Integer |
1 |
Number of notes. |
Contour Descriptors¶
Description: Calculates various contour descriptors. First, Huron contour in three different varieties is supported: Code, Reduced Code and Numeric. Huron contour is calculated on a sequence of pitches by size comparison of the first pitch to the mean inner pitch (MIP, i.e., mean of MIDI pitches except the first and last element), and the MIP to the last pitch. There are three possible relations between two numbers: Greater (>), equal (=), or less (<). Combining two comparisons gives a set of nine different contour values: descending (>>), descending-horizontal (>=), concave (<>), horizontal-descending (=>), horizontal (==), horizontal-ascending (=<), convex (><), ascending-horizontal (<=), and ascending (<<). Short mnemonic codes for them are
desc
,desc-hor
,concave
,hor-desc
,hor
,hor-asc
,convex
,asc-hor
,asc
. The reduced code is a mapping of the mixed-horizontal category to their non-horizontal partner, i.e.,desc-hor
andhor-desc
becomedesc
andasc-hor
andhor-asc
becomeasc
, which leaves 5 categories. The numeric values range from -4 to 4 in the same order. (Cf. Huron (1994), The Melodic Arch in Western Folksongs.). Second, gradient countour measures the pitch gradient between extreme points. Third, Abesser contour is similar to Huron contour, but uses more stable estimators and is more suitable for longer sequences.Feature Definition File Name:
CONTOUR
Category:
Contour
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
1 |
Huron contour codes. One of |
|
String |
1 |
Reduced Huron contour codes. One of |
|
Integer [-4:+4] |
1 |
Numeric value of Huron contour. |
|
Real |
Var |
List of gradients between extreme values of note track. |
|
String |
1 |
Abesser contour codes. One of |
|
String |
1 |
Reduced Abesser contour codes. One of |
|
Integer [-4:+4] |
1 |
Numeric value for Abesser contour. |
Interval¶
Parson’s Code Single Densities¶
Description: Normalized histogram densities, mode, entropy and Zipf coefficient for Parson’s code. Values: down (-1), repeat/unisone (0), up (+1).
Feature Definition File Name:
PARSON_FEATURES
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency for down (-1). |
|
Real |
1 |
Relative frequency for repeat (0). |
|
Real |
1 |
Relative frequency for up (+1). |
|
Real |
Var |
Mode of Parson’s Code distribution (can be list!). |
|
Real |
1 |
Normalized entropy of Parson’s Code distribution. |
|
Real |
1 |
Zipf coefficient of Parson’s Code distribution. |
Semitone Interval Single Features¶
Description: Calculates set of basic statistical sample descriptors (mean, median, mode, min, max, range, var, std) for interval and absolute interval distribution.
Feature Definition File Name:
INT_FEATURES
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean of interval distribution. |
|
Real |
1 |
Median of interval distribution. |
|
Real |
1 |
Variance of of interval distribution. |
|
Real |
1 |
Standard deviation of interval distribution. |
|
Real |
1 |
Minimum of of interval distribution. |
|
Real |
1 |
Maximum of of interval distribution. |
|
Real |
1 |
Range of of interval distribution. |
|
Real |
Var |
Mode of interval distribution (can be list!). |
|
Real |
1 |
Zipf coefficient of interval distribution. |
|
Real |
1 |
Entropy of interval distribution (bits). |
|
Real |
1 |
Mean of absolute interval distribution. |
|
Real |
1 |
Median of absolute interval distribution. |
|
Real |
1 |
Variance of absolute interval distribution. |
|
Real |
1 |
Standard deviation of interval distribution. |
|
Real |
1 |
Minimum of absolute interval distribution. |
|
Real |
1 |
Maximum of absolute interval distribution. |
|
Real |
1 |
Range of absolute interval distribution. |
|
Real |
1 |
Mode of interval absolute distribution (can be list!). |
|
Real |
1 |
Zipf coefficient of absolute interval distribution. |
|
Real |
1 |
Entropy of absolute interval distribution (bits). |
|
Real |
1 |
Sum of absolute intervals (= distance moved in pitch space). |
Parson’s Code Distribution (“Contour”)¶
Description: Histogram and raw vector of Parson’s Code (contour). Values: down (-1), repeat/unisone (0), up (+1).
Feature Definition File Name:
PARSON_HIST
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [-1:+1] |
N |
Raw output (Parson’s code values of all melody notes). |
|
Integer [-1:+1] |
3 |
Histogram bins for Parson’s code. |
|
Real |
3 |
Histogram bin counts (abs. frequencies). |
Semitone Interval Distribution¶
Description: Histogram and raw vector of semitone intervals, either limited to the range of -20 to 20 semitones or full range (only raw and with appended zero element).
Feature Definition File Name:
INT_HIST
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [-127:+127] |
N-1 |
ist of semitone intervals. |
|
Integer [-20:+20] |
41 |
Bins of semitone intervals [-20:20]. |
|
Integer |
41 |
Absolute frequencies of semitone intervals. |
|
Integer [-127:+127] |
N |
List of semitone intervals with appended 0. |
Fuzzy Interval Distribution (Refined Contour)¶
Description: Histogram and raw vector of fuzzy intervals (aka refined contour). Values: big jump up/down (+/-4), jump up/down (+/-3), leap up/down (+/-1), step up/down (+/-1), repetition (0).
Feature Definition File Name:
FUZZYINT_HIST
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [-4:4] |
N-1 |
Raw output (Fuzzy interval values over the full melody). |
|
Integer [-4:4] |
5 |
Histogram bins. |
|
Real |
5 |
Histogram bin counts (abs. frequencies). |
Fuzzy Interval (Refined Contour) Single Features¶
Description: Histogram counts, mode, entropy and Zipf coefficient of distribution of all 9 fuzzy interval classes (FUZZYINT). Values: big jump up/down (+/-4), jump up/down** (+/-3), leap up/down (+/-1), step up/down (+/-1), repetition (0).
Feature Definition File Name:
FUZZYINT_FEATURES
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency big jump down (-4). |
|
Real |
1 |
Relative frequency jump down (-3). |
|
Real |
1 |
Relative frequency leap down ( -2). |
|
Real |
1 |
Relative frequency step down (-1). |
|
Real |
1 |
Relative frequency repetition (0). |
|
Real |
1 |
Relative frequency step up (+1). |
|
Real |
1 |
Relative frequency leap up (+2). |
|
Real |
1 |
Relative frequency jump up (+3). |
|
Real |
1 |
Relative frequency big jump up (+4). |
|
Real |
Var |
Mode of FUZZYINT distribution (can be list!). |
|
Real |
1 |
Zipf coefficient of FUZZYINT distribution. |
|
Real |
1 |
Normalized entropy of FUZZYINT distribution. |
Interval Class Distribution¶
Description: Histogram and raw vector of intervals classes. Interval classes are minimal differences between pitches classes (i.e., length of shortest path on the circle) and have values from 0 to 6.
Feature Definition File Name:
INTCLASS_HIST
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0:+6] |
N-1 |
List of interval classes. |
|
Integer [0:6] |
7 |
Bins of Interval classes [0:6]. |
|
Integer |
7 |
Absolute frequencies of interval classes. |
|
Integer [0:+6] |
N |
List of interval classes with appended 0. |
Interval Class Single Features¶
Description: Normalized histogram count, and basic statistics of interval class distribution. Interval classes are minimal differences between pitches classes (i.e., length of shortest path on the circle) and have values from 0 to 6.
Feature Definition File Name:
INTCLASS_FEATURES
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean of interval class distribution. |
|
Real |
1 |
Median of interval class distribution. |
|
Real |
1 |
Variance of of interval class distribution. |
|
Real |
1 |
Standard deviation of interval class distribution. |
|
Real |
1 |
Minimum of of interval class distribution. |
|
Real |
1 |
Maximum of of interval class distribution. |
|
Real |
1 |
Range of of interval class distribution. |
|
Real |
Var |
Mode of interval class distribution (can be list!). |
|
Real |
1 |
Zipf coefficient of interval class distribution. |
|
Real |
1 |
Entropy of interval class distribution (bits). |
|
Real |
1 |
Relative frequency IC = 0. |
|
Real |
1 |
Relative frequency IC = 1. |
|
Real |
1 |
Relative frequency IC = 2. |
|
Real |
1 |
Relative frequency IC = 3. |
|
Real |
1 |
Relative frequency IC = 4. |
|
Real |
1 |
Relative frequency IC = 5. |
|
Real |
1 |
Relative frequency IC = 6. |
Parson’s Code Up/Down Ratio¶
Description: Ratio of number of ascending and descending intervals in a melody (ignoring unisones).
Feature Definition File Name:
PARSON_UP_DOWN_RATIO
Category:
Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0,1] |
1 |
Ratio of ascending to descending intervals. |
MLA¶
MLA Exports¶
Description: Exports several mid-level analysis (MLA) related features (section event-based).
Feature Definition File Name:
MLA_RAW
Category:
MLA
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
Var |
Raw list of mid-level units. |
|
String |
Var |
Raw list of main types of mid-level units. |
|
String |
Var |
Raw list of full types of mid-level units. |
|
String |
Var |
Raw list of back references for mid-level units. |
|
String |
Var |
Raw list of glue operators of mid-level units. |
|
String |
Var |
Raw list of modifiere for mid-level units. |
|
String |
Var |
Raw list of idea length as number of notes. |
|
String |
Var |
Raw list of idea duration in secs onset-offset. |
|
String |
Var |
Raw list of idea duration in decimal bar time. |
|
String |
Var |
Raw list of main directions of mid-level units. |
Metadata¶
Solo Metadata¶
Description: Export all metadata for Jazzomat solos (currently only a subset).
Feature Definition File Name:
GENERAL_SOLO_METADATA
Category:
Metadata
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
1 |
Performer. |
|
String |
1 |
Title. |
|
String |
1 |
Add-on for title. |
|
String |
1 |
Running number of a solo in a piece. |
|
String |
1 |
Full title (including solo part and title addon). |
|
String |
1 |
Instrument of the soloist. One of |
|
String |
1 |
Style of piece. One of |
|
String |
1 |
Avg. Tempo (bpm). |
|
String |
1 |
Tempo class. One of |
|
String |
1 |
Basic rhythmic feel/groove style. One of |
|
String |
1 |
Key or tonal center of the piece. |
|
String |
1 |
Filename of originating SV project file. |
|
String |
1 |
Status of transcription. One of |
|
String |
1 |
Composer of the piece. |
|
String |
1 |
Tonality type of the composition. One of |
|
String |
1 |
Genre of the composition. One of |
|
String |
1 |
Form of the composition. |
|
String |
1 |
Signature(s) of solo. |
|
String |
1 |
Music Brainz ID. |
|
String |
1 |
Recording date(s) of containing record. |
|
String |
1 |
Recording year of containing record. |
|
String |
1 |
Bibliography of record. |
|
String |
1 |
Title of containing record. |
|
String |
1 |
Record label of containing record. |
|
String |
1 |
Start time of solo in full track. |
|
String |
1 |
End time of solo in full track. |
|
String |
1 |
Line-up of track. |
|
String |
1 |
Chord changes as compact string. |
|
String |
1 |
Number of full choruses in the solo. |
|
String |
1 |
Underlying harmony template of composition. |
|
String |
1 |
Start/End of solo in surrounding track. |
EsAC Metadata¶
Description: Export all metadata for EsAC tunes.
Feature Definition File Name:
GENERAL_ESAC_METADATA
Category:
Metadata
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
1 |
Collection. |
|
String |
1 |
Orignal EsAC title. |
|
String |
1 |
Region of origin. |
|
String |
1 |
Annotated key (tonic). |
|
Integer (2,4,8,16,32) |
1 |
Annotated minimal rhythmic unit. |
|
String |
1 |
Annotated signature (possible more than one, space separate, or FREE). |
|
String |
1 |
Annotated key (tonic). |
|
String |
1 |
Annotated (ritual) function. |
|
String |
1 |
Comment section. |
|
String |
1 |
CNR section (with yet unknown meaning!). |
|
String |
1 |
Tunefamily based on EsAC-Ids. |
|
String |
1 |
Annotated lyrics. |
|
String |
1 |
Original melody code. |
Meter¶
Metrical Circle Map Markov Distributions¶
Description: Histogram for Metrical Circle Map (N=48) unigrams and bigrams (occurrence and transition probabilites). See this document or here for more details.
Feature Definition File Name:
MCM_HIST
Category:
Meter
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:47] |
48 |
Histogram bins (MCM:48 values). |
|
Real [0,1] |
48 |
Histogram bin densities (MCM:48 values). |
|
Array of Integer |
Var |
Histogram bins (all unique MCM bigrams). |
|
Real [0,1] |
Var |
Histogram bin densities (all unique MCM bigrams). |
Metrical Circle Map Markov Features¶
Description: Histogram for Metrical Circle Map (N=48) unigrams and bigrams (occurrence and transition probabilites, entropies and Zipf coefficients). See this document or here for more details.
Feature Definition File Name:
MCM_FEATURES
Category:
Meter
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Angle of circular mean of MCM distribution. |
|
Real |
1 |
Length of circular mean of MCM distribution. |
|
Real |
1 |
Circular variance of MCM distribution (=1-length of circular mean). |
|
Real |
1 |
Circular standard deviation of MCM distribution. |
|
Real |
1 |
Circular dispersion of MCM distribution. |
|
Real |
Var |
Mode of MCM distribution (can be list!). |
|
Real |
1 |
Zipf coefficient of MCM distribution. |
|
Real |
1 |
Normalized entropy of MCM distribution. |
Meter Raw Exports¶
Description: Exports several meter related features (event-based).
Feature Definition File Name:
METER_RAW
Category:
Meter
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
N |
Metrical positions (in MPD-notation) of events. |
|
Integer |
N |
Event bar numbers. |
|
Integer |
N |
Event beat numbers. |
|
Integer |
N |
Event tatum positions. |
|
Integer |
N |
Events durations measured in current tatums (handle with care!). |
|
Integer [0:2] |
N |
Metrical weight of event (0 for any subbeat event, 1 for metrical weak beat events, 2 for metrical strong events). |
|
Integer [0:47] |
N |
Position in metrical circle map with N=48 divisions. |
|
Integer [0:8] |
N-1 |
Bigrams of metrical weights coded with ternary numbers. Let , be the weights of notes . Then the metrical weight transition value is . |
|
Integer |
N |
Event-based periods. |
|
Integer |
N |
Event-based beat divisions. |
|
Integer [0:1] |
N |
Marker if event is syncopated (0=not syncopated, 1=syncopated). |
|
Real [0,1] |
1 |
Ratio of syncopated events to non-syncopated events. |
|
Real [0,1] |
1 |
Combined metric complexity after Frieler, in preparation. |
|
Real [0,1] |
1 |
Division-based metric complexity after Frieler, in preparation. |
|
Real [0,1] |
1 |
Compression-based metric complexity after Frieler, in preparation. |
Pitch¶
Pitch Features¶
Description: Calculates set of basic statistical sample descriptors (mean, median, mode, min, max, range, var, std).
Feature Definition File Name:
PITCH_FEATURES
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean of PITCH distribution. |
|
Real |
1 |
Median of PITCH distribution. |
|
Real |
1 |
Variance of of PITCH distribution. |
|
Real |
1 |
Standard deviation of PITCH distribution. |
|
Real |
1 |
Minimum of of PITCH distribution. |
|
Real |
1 |
Maximum of of PITCH distribution. |
|
Real |
1 |
Range of of PITCH distribution. |
|
Real |
Var |
Mode of PITCH distribution (can be list!). |
|
Real |
1 |
Zipf coefficient of PITCH distribution. |
|
Real |
1 |
Entropy of PITCH distribution (bits). |
Tessitura Normalized Pitch¶
Description: Raw vector of scaled MIDI pitches with respect to pitch range.
Feature Definition File Name:
TESS_NORM_PITCH
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0, 1] |
N |
Tessatura-scaled pitch values. |
Chordal Diatonic Pitch Class Distribution¶
Description: Histogram and raw vector of chordal diatonic pitch classes (CDPC).
Feature Definition File Name:
CDPC_HIST
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer |
N |
Raw output of note-wise CDPC values. |
|
String |
Var |
Histogram bins (all unique CDPC values occurring in a melody). |
|
Real |
Var |
Histogram bin counts (absolute frequencies). |
|
Integer |
N |
Raw output of note-wise CDPC values including NC values. |
Number of Different Pitch Classes.¶
Description: Number of different pitch classes.
Feature Definition File Name:
PC_NUM_UNIQUE_PC
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer |
1 |
Average number of different pitch classes. |
Tonal Pitch Class Single Features¶
Description: Histogram counts of all 12 TPC values (normalized to density), as well as basic circular statistics (cf. https://en.wikipedia.org/wiki/Directional_statistics ).
Feature Definition File Name:
TPC_FEATURES
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency TPC = 0. |
|
Real |
1 |
Relative frequency TPC = 1. |
|
Real |
1 |
Relative frequency TPC = 2. |
|
Real |
1 |
Relative frequency TPC = 3. |
|
Real |
1 |
Relative frequency TPC = 4. |
|
Real |
1 |
Relative frequency TPC = 5. |
|
Real |
1 |
Relative frequency TPC = 6. |
|
Real |
1 |
Relative frequency TPC = 7. |
|
Real |
1 |
Relative frequency TPC = 8. |
|
Real |
1 |
Relative frequency TPC = 9. |
|
Real |
1 |
Relative frequency TPC = 10. |
|
Real |
1 |
Relative frequency TPC = 11. |
|
Real |
1 |
Angle of circular mean of TPC distribution. |
|
Real |
1 |
Length of circular mean of TPC distribution. |
|
Real |
1 |
Circular variance of TPC distribution (=1-length of circular mean). |
|
Real |
1 |
Circular standard deviation of TPC distribution. |
|
Real |
1 |
Circular dispersion of TPC distribution. |
|
Real |
1 |
Normalized entropy of TPC distribution. |
Extended Chordal Diatonic Pitch Class Distribution¶
Description: Histogram and raw vector of extended multi-digit chordal diatonic pitch classes (CDPCX).
Feature Definition File Name:
CDPCX_HIST
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer |
N |
Raw output of note-wise CDPCX values. |
|
Integer |
N |
Histogram bins (all unique CDPCX values occurring in a melody). |
|
Real |
N |
Histogram bin counts (abs. frequencies). |
|
Integer |
N |
Raw output of note-wise CDPCX values including NC values. |
Chordal Diatonic Pitch Class Features¶
Description: Normalized histogram counts, mode, entropy and Zipf coefficient of CDPC distribution.
Feature Definition File Name:
CDPC_FEATURES
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency CDPC = ‘1’. |
|
Real |
1 |
Relative frequency CDPC = ‘2’. |
|
Real |
1 |
Relative frequency CDPC = ‘3’. |
|
Real |
1 |
Relative frequency CDPC = ‘4’. |
|
Real |
1 |
Relative frequency CDPC = ‘5’. |
|
Real |
1 |
Relative frequency CDPC = ‘6’. |
|
Real |
1 |
Relative frequency CDPC = ‘7’. |
|
Real |
1 |
Relative frequency CDPC = ‘T’. |
|
Real |
1 |
Relative frequency CDPC = ‘B’. |
|
Real |
1 |
Relative frequency CDPC = ‘L’. |
|
Real |
1 |
Relative frequency CDPC = ‘<’. |
|
Real |
1 |
Relative frequency CDPC = ‘>’. |
|
Real |
Var |
Mode of CDPC distribution (can be list!). |
|
Real |
1 |
Normalized entropy of CDPC distribution. |
|
Real |
1 |
Zipf coefficient of CDPC distribution. |
Pitch Class Single Features¶
Description: Histogram counts of all 12 PC values (normalized to density), as well as basic circular statistics (cf. https://en.wikipedia.org/wiki/Directional_statistics ).
Feature Definition File Name:
PC_FEATURES
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency PC = C. |
|
Real |
1 |
Relative frequency PC = C#. |
|
Real |
1 |
Relative frequency PC = D. |
|
Real |
1 |
Relative frequency PC = Eb. |
|
Real |
1 |
Relative frequency PC = E. |
|
Real |
1 |
Relative frequency PC = F. |
|
Real |
1 |
Relative frequency PC = F#. |
|
Real |
1 |
Relative frequency PC = G. |
|
Real |
1 |
Relative frequency PC = Ab. |
|
Real |
1 |
Relative frequency PC = A. |
|
Real |
1 |
Relative frequency PC = Bb. |
|
Real |
1 |
Relative frequency PC = B. |
|
Real |
1 |
Angle of circular mean of PC distribution. |
|
Real |
1 |
Length of circular mean of PC distribution. |
|
Real |
1 |
Circular variance of PC distribution (=1-length of circular mean). |
|
Real |
1 |
Circular standard deviation of PC distribution. |
|
Real |
1 |
Circular dispersion of PC distribution. |
|
Real |
1 |
Normalized entropy of PC distribution. |
Average Number of Pitch Classes per Phrase¶
Description: Average number of different pitch classes per phrase.
Feature Definition File Name:
PC_AV_NUM_UNIQUE_PC_PHRASES
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0:11] |
1 |
Average number of different pitch classes per phrase. |
Extended Chordal Diatonic Pitch Class Single Features¶
Description: Normalized histogram counts, mode, entropy and Zipf coefficient of CDPCX distribution.
Feature Definition File Name:
CDPCX_FEATURES
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency CDPCX = ‘1’. |
|
Real |
1 |
Relative frequency CDPCX = ‘2’. |
|
Real |
1 |
Relative frequency CDPCX = ‘3’. |
|
Real |
1 |
Relative frequency CDPCX = ‘4’. |
|
Real |
1 |
Relative frequency CDPCX = ‘5’. |
|
Real |
1 |
Relative frequency CDPCX = ‘6’. |
|
Real |
1 |
Relative frequency CDPCX = ‘7’. |
|
Real |
1 |
Relative frequency CDPCX = ‘T’. |
|
Real |
1 |
Relative frequency CDPCX = ‘B’. |
|
Real |
1 |
Relative frequency CDPCX = ‘L’. |
|
Real |
1 |
Relative frequency CDPCX = ‘<’. |
|
Real |
1 |
Relative frequency CDPCX = ‘>’. |
|
Real |
1 |
Relative frequency CDPCX = ‘-‘ (b2). |
|
Real |
1 |
Relative frequency CDPCX = ‘%’ (b6). |
|
Real |
Var |
Mode of CDPCX distribution (can be list!). |
|
Real |
1 |
Normalized entropy of CDPCX distribution. |
|
Real |
1 |
Zipf coefficient of CDPCX distribution. |
Tonal Pitch Class Distribution¶
Description: Histogram and raw vector of tonal pitch classes (TPC).
Feature Definition File Name:
TPC_HIST
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer |
N |
Raw output of note-wise TPC values. |
|
Integer [0:11] |
12 |
Histogram bins (all unique TPC values occurring in a melody). |
|
Integer |
12 |
Histogram bin count values (abs. frequencies). |
Pitch Class Distribution¶
Description: Histogram and raw vector of pitch classes.
Feature Definition File Name:
PC_HIST
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0:11] |
N |
Raw output (values for all melody notes). |
|
Integer [0:11] |
12 |
Histogram bins. |
|
Real |
12 |
Histogram bin count (abs. frequencies). |
Chordal Pitch Class Feature¶
Description: Normalized histogram count, and basic circular statistics (cf. https://en.wikipedia.org/wiki/Directional_statistics ) of CPC distribution.
Feature Definition File Name:
CPC_FEATURES
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency CPC = 0. |
|
Real |
1 |
Relative frequency CPC = 1. |
|
Real |
1 |
Relative frequency CPC = 2. |
|
Real |
1 |
Relative frequency CPC = 3. |
|
Real |
1 |
Relative frequency CPC = 4. |
|
Real |
1 |
Relative frequency CPC = 5. |
|
Real |
1 |
Relative frequency CPC = 6. |
|
Real |
1 |
Relative frequency CPC = 7. |
|
Real |
1 |
Relative frequency CPC = 8. |
|
Real |
1 |
Relative frequency CPC = 9. |
|
Real |
1 |
Relative frequency CPC = 10. |
|
Real |
1 |
Relative frequency CPC = 11. |
|
Real |
1 |
Angle of circular mean of CPC distribution. |
|
Real |
1 |
Length of circular mean of CPC distribution. |
|
Real |
1 |
Circular variance of CPC distribution (=1-length of circular mean). |
|
Real |
1 |
Circular standard deviation of CPC distribution. |
|
Real |
1 |
Circular dispersion of CPC distribution. |
|
Real |
1 |
Zipf coefficient of CPC distribution. |
|
Real |
1 |
Normalized entropy of CPC distribution. |
Chordal Pitch Class Distribution¶
Description: Histogram and raw vector of chordal pitch classes (CPC).
Feature Definition File Name:
CPC_HIST
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0:11] |
N |
Raw output (CPC values of all melody notes). |
|
Integer [-1:11] |
N |
Raw output (CPC values of all melody notes, including undefined CPCs, set to -1). |
|
Integer [0:11] |
12 |
Histogram bins. |
|
Real |
12 |
Histogram bin count (absolute frequencies). |
Pitch Distribution¶
Description: Histogram and raw vector of raw MIDI pitches.
Feature Definition File Name:
PITCH_HIST
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer [0:127] |
N |
Raw output (values of all melody notes). |
|
Integer [0:127] |
128 |
Histogram bins. |
|
Real |
128 |
Histogram bin counts (abs. frequencies). |
Chordal Pitch Class Circle Map Features (Experimental)¶
Description: Histogram for Chordal Pitch Class Circle Map (N=12) unigrams and bigrams (occurrence and transition probabilites).
Feature Definition File Name:
CPCCM_HIST
Category:
Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:11] |
12 |
Histogram bins (CPCCM:12 values). |
|
Real [0,1] |
12 |
Histogram bin densities (CPCCM:12 values). |
|
Array of Integer |
Var |
Histogram bins (all unique CPCCM bigrams). |
|
Real [0,1] |
Var |
Histogram bin densities (all unique CPCCM bigrams). |
Rhythm¶
Variability Indices (Duration)¶
Description: Normalized pairwise variability index, which measures the mean contrast of adjacents durations. The formula is . It is the mean of the absolute value of ratios of duration difference to duration mean. This index originated in linguistic and is sometimes used by rhythm researchers as a variability measure. The coefficient of variation is another common index for variability. It is defined as the ratio of standard deviation and mean. A simple biased estimator is the corresponding ratio of sample standard deviation and sample mean. The coefficient of variation is only sensical for ratio scales, which applies to durations since they only assum positive values. The formula is: .
Feature Definition File Name:
VI_DUR
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0,1] |
1 |
Normalized pairwise variability index. |
|
Real |
1 |
Coefficient of variation. |
Microtiming Features¶
Description: Features related to micro-timing in a solo.
Feature Definition File Name:
GENERAL_MICROTIMING
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
Var |
Swing ratios of beats with a binary sub-division as the ratio duration of the first to the second eighth in a beat. Straight binary eigths have a value of 1 (1:1), whereas triplet eighths get 2 and dotted eighths + sixteenth get 3:1. |
|
Real |
1 |
Mean value of swing ratios. |
|
Real |
1 |
Median of swing ratios. |
|
Real |
1 |
Standard deviation of swing ratio. |
|
Real |
Var |
Swing shapes are the loudness difference between first and second eighth in a beat with binary subdivision. |
|
Real |
1 |
Mean value of swing shapes. |
|
Real |
1 |
Median of swing shapes. |
|
Real |
1 |
Standard deviation of swing shapes. |
|
Real |
1 |
Number of fully occupied binary beats. |
|
Real |
1 |
Proportion of fully occupied binary beats. |
|
Real |
N |
Nominal metrical onsets according to metrical annotation and beat track. |
|
Real |
N |
Difference of onsets to nominal metrical onsets. |
|
Real |
N |
Absolute difference of onsets to nominal metrical onsets. |
Duration Ratios¶
Description: Raw vector of duration ratios as well as duration ratio classification (-1: shorter, 0: equal, +1: longer) using thresholds 1.49 and .81.
Feature Definition File Name:
DURRATIO_HIST
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Float |
N-2 |
Raw output of duration ratios. |
|
Integer [-1:1] |
N-2 |
Classified duration ratios. |
Inter-onset Interval Class Single Features¶
Description: Normalzed histogram counts, mode,entropy and Zipf coefficent for all 5 duration classes (very short: -2, short: -1, medium: 0, long: 1, very long: 2). Reference time-span is either local beat duration (“relative mode”) or 0.5 sec (“absolute mode”).
Feature Definition File Name:
IOICLASS_FEATURES
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency very short (absolute mode). |
|
Real |
1 |
Relative frequency short (absolute mode). |
|
Real |
1 |
Relative frequency medium (absolute mode). |
|
Real |
1 |
Relative frequency long (absolute mode). |
|
Real |
1 |
Relative frequency very long (absolute mode). |
|
Real |
1 |
Relative frequency very short (relative mode). |
|
Real |
1 |
Relative frequency short (relative mode). |
|
Real |
1 |
Relative frequency medium (relative mode). |
|
Real |
1 |
Relative frequency long (relative mode). |
|
Real |
1 |
Relative frequency very long (relative mode). |
|
Real |
Var |
Mode of absolute inter-onset interval class distributions (can be list). |
|
Real |
1 |
Entropy of absolute inter-onset interval class distributions. |
|
Real |
1 |
Zipf coefficient of absolute durations class distributions. |
|
Real |
Var |
Mode of relative inter-onset interval classdistributions (can be list). |
|
Real |
1 |
Entropy of relative inter-onset interval class distributions. |
|
Real |
1 |
Zipf coefficient of relative inter-onset interval class distributions. |
Inter-onset Interval Class Distribution¶
Description: Histogram and raw vector of IOI classes (very short: -2, short: -1, medium: 0, long: 1, very long: 2). Reference time-span is either local beat duration (“relative mode”) or 0.5 sec (“absolute mode”).
Feature Definition File Name:
IOICLASS_HIST
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer |
N |
Raw output of IOI classes of all melody notes (absolute mode). |
|
Integer [-2:2] |
5 |
Histogram bin values (absolute mode). |
|
Integer |
5 |
Histogram bin counts (absolute mode). |
|
Integer |
N |
Raw output of IOI classes of all melody notes (relative mode). |
|
Integer [-2:2] |
5 |
Histogram bin values (relative mode). |
|
Integer |
5 |
Histogram bin counts (relative mode). |
Event Densities¶
Description: Events per second and per bar.
Feature Definition File Name:
GENERAL_EVENT_DENSITY
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Event density (notes per seconds). |
|
Real |
1 |
Total duration in seconds. |
|
Real |
1 |
Event density (notes per bar). WARNING: Might not be working properly with meter changes. |
|
Real |
1 |
Total duration in fractional bar units (e.g., 6 Quarter notes in 4/4 timing are 1.5 bars long). |
|
Integer |
1 |
Number of events. |
Duration Class Single Features¶
Description: Normalized histogram counts, mode, entropy and Zipf cofficient for distribution if all 5 duration classes (very short: -2, short: -1, medium: 0, long: 1, very long: 2) in relative and absolute mode. Reference duration is either local beat duration (‘relative mode’) or 0.5 sec (‘absolute mode’).
Feature Definition File Name:
DURCLASS_FEATURES
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Relative frequency very short. |
|
Real |
1 |
Relative frequency short. |
|
Real |
1 |
Relative frequency medium. |
|
Real |
1 |
Relative frequency long. |
|
Real |
1 |
Relative frequency very long. |
|
Real |
1 |
Relative frequency very short. |
|
Real |
1 |
Relative frequency short. |
|
Real |
1 |
Relative frequency medium. |
|
Real |
1 |
Relative frequency long. |
|
Real |
1 |
Relative frequency very long. |
|
Real |
Var |
Mode of absolute duration class distributions (can be list). |
|
Real |
1 |
Entropy of absolute duration class distributions. |
|
Real |
1 |
Zipf coefficient of absolute duration class distributions. |
|
Real |
Var |
Mode of relative duration class distributions (can be list). |
|
Real |
1 |
Entropy of relative duration class distributions. |
|
Real |
1 |
Zipf coefficient of relative duration class distributions. |
Duration Class Distribution¶
Description: Histogram and raw vector of duration classes (very short: -2, short: -1, medium: 0, long: 1, very long: 2). Reference duration is either local beat duration (‘relative mode’) or 0.5 sec (‘absolute mode’).
Feature Definition File Name:
DURCLASS_HIST
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Integer |
N |
Raw output of duration classes (absolute mode). |
|
Integer [-2:2] |
5 |
Histogram bin values (absolute mode). |
|
Integer |
5 |
Histogram bin counts (abs. frequencies, absolute mode). |
|
Integer |
N |
Raw output of duration classes (relative mode). |
|
Integer [-2:2] |
5 |
Histogram bin values (relative mode). |
|
Integer |
5 |
Histogram bin counts (abs. frequencies, relative mode). |
Variability Indices (IOI)¶
Description: Normalized pairwise variability index, which measures the mean contrast of adjacents inter-onset intervals. The formula is . It is the mean of the absolute value of ratios of IOI difference to IOI mean. This index originated in linguistic and is sometimes used by rhythm researchers as a variability measure. The coefficient of variation is another common index for variability. It is defined as the ratio of standard deviation and mean. A simple biased estimator is the corresponding ratio of sample standard deviation and sample mean. The coefficient of variation is only sensical for ratio scales, which applies to durations since they only assum positive values. The formula is: .
Feature Definition File Name:
VI_IOI
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0,1] |
1 |
Normalized pairwise variability index. |
|
Real |
1 |
Coefficient of variation. |
IOI Ratios¶
Description: Raw vector of IOI ratios as well as IOI ratio classification (-1: shorter, 0: equal, +1: longer) using thresholds 1.49 and .81.
Feature Definition File Name:
IOIRATIO_HIST
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Float |
N-2 |
Raw output of IOI ratios. |
|
Integer [-1:1] |
N-2 |
Classified IOI ratios. |
Tempo Features¶
Description: Statistical descriptors of tempo distribution (mean, stddev, min, max, range) for inter-beat intervals (IBI) measured in seconds or BPM counts (beats per minute). BPM=60/IBI.
Feature Definition File Name:
GENERAL_TEMPO
Category:
Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean tempo of a melody measured in seconds (average IBI). |
|
Real |
1 |
Standard devitation of IBI distribution in seconds. |
|
Real |
1 |
Range of IBI distribution (max-min) in seconds. |
|
Real |
1 |
Minimum of IBI distribution (min) in seconds. |
|
Real |
1 |
Maximum of IBI distribution (max) in seconds. |
|
Real |
1 |
Mean tempo of a melody measured in BPM (beats per minute, BPM = 60/IBI). |
|
Real |
1 |
Standard devitation of BPM distribution. |
|
Real |
1 |
Range (max.min) of BPM distribution. |
|
Real |
1 |
Minimum of BPM distribution. |
|
Real |
1 |
Maximum of BPM distribution. |
|
Real |
Var |
Raw annotated beat positions. |
Sequence/Interval¶
Interval Class Bigrams¶
Description: Histogram features for intervals classes (IC) bigrams: all unique IC bigrams and all IC bigrams occurring at least twice.
Feature Definition File Name:
INTCLASS_2GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer |
Var |
Histogram bins (all unique IC bigrams). |
|
Integer |
Var |
Histogram bin counts (all unique IC bigrams). |
|
Array of Integer |
Var |
Histogram bins (all unique IC bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin counts (all unique IC bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Fuzzy Interval Trigrams¶
Description: Histogram features for fuzzy intervals (FUZZYINT, also known as refined contour) trigrams: all unique FUZZYINT trigrams and all FUZZYINT trigrams occurring at least twice.
Feature Definition File Name:
FUZZYINT_3GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [-4:4] |
Var |
Histogram bins (all unique FUZZYINT trigrams). |
|
Integer |
Var |
Histogram bin count (all unique FUZZYINT trigrams). |
|
Array of Integer [-4:4] |
Var |
Histogram bins (all unique FUZZYINT trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique FUZZYINT trigrams occurring at least twice). |
Average Run-lengths of arpeggios¶
Description: Run-lengths and means of arpeggios with intervals of size +3/+4 or -3/-4 (fuzzy interval classes leap up/down.).
Feature Definition File Name:
ARPEGGIOS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean length of ascending arpeggios. |
|
Real |
1 |
Mean length of descending arpeggios. |
|
Real |
1 |
Mean length of arpeggios (mixed up/down thirds). |
Average Run-lengths of Chromatic passages¶
Description: Run-lengths and means of chromatic sequences with intervals of size +1 or -1.
Feature Definition File Name:
STEP_SEQUENCES
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean length of ascending step passages. |
|
Real |
1 |
Mean length of descending step passages. |
|
Real |
1 |
Mean length of step passages (up and down steps). |
Fuzzy Interval Bigrams¶
Description: Histogram features for fuzzy intervals (FUZZYINT, also known as refined contour) bigrams: all unique FUZZYINT bigrams and all FUZZYINT bigrams occurring at least twice.
Feature Definition File Name:
FUZZYINT_2GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [-4:4] |
Var |
Histogram bins (all unique FUZZYINT bigrams). |
|
Integer |
Var |
Histogram bin count (all unique FUZZYINT bigrams). |
|
Array of Integer [-4:4] |
Var |
Histogram bins (all unique FUZZYINT bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique FUZZYINT bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Average Run-lengths of Chromatic Passages¶
Description: Run-lengths and means of chromatic sequences with intervals of size +1 or -1.
Feature Definition File Name:
CHROMATIC_SEQUENCES
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean length of ascending chromatic passages. |
|
Real |
1 |
Mean length of descending chromatic passages. |
|
Real |
1 |
Mean length of chromatic passages (up and down combined). |
Parson’s Code Bigrams¶
Description: Histogram features for Parson’s Code (PARSON) bigrams: all unique PARSON bigrams and all PARSON bigrams occurring at least twice.
Feature Definition File Name:
PARSON_2GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [-1:+1] |
Var |
Histogram bins (all unique PARSON bigrams). |
|
Integer |
Var |
Histogram bin count (all unique PARSON bigrams). |
|
Array of Integer [-1:+1] |
Var |
Histogram bins (all unique PARSON bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique PARSON bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Percentage of Chromatic Passages¶
Description: Computes the percentage of chromatic passages (interval sequences with values +1 or -1 with at least three notes), in the set of all passages (sequences of constant interval).
Feature Definition File Name:
INT_CHROMATIC_SEQUENCES_RATIO
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Percentage of of chromatic passages. |
Average Run-length of Parson’s Code Segments¶
Description: Means of run-length of segments of notes with constant interval direction.
Feature Definition File Name:
PARSON_CONST_DIRECTION_AV_LEN
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean run-length of unisone sequences. |
|
Real |
1 |
Mean run-length of strictly ascending sequences. |
|
Real |
1 |
Mean run-length of strictly descending sequences. |
Average Run-length of Chromatic Passages¶
Description: Mean run-length of sequences with intervals of size +1 or -1.
Feature Definition File Name:
INT_CHROMATIC_SEQUENCES_AV_LEN
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean run-length of chromatic passages. |
Average Run-lengths of Parsons Sequences¶
Description: Run-lengths and means of sequences of all (Parson’s) intervals directions.
Feature Definition File Name:
PARSON_SEQUENCES
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean length of ascending passages. |
|
Real |
1 |
Mean length of descending passages. |
|
Real |
1 |
Mean length of repetitions. |
Semitone Interval Trigrams¶
Description: Histogram features for semitone intervals (INT) trigrams: all unique INT trigrams and all INT trigrams occurring at least twice.
Feature Definition File Name:
INT_3GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer |
Var |
Histogram bins (all unique INT trigrams). |
|
Integer |
Var |
Histogram bin counts (all unique INT trigrams). |
|
Array of Integer |
Var |
Histogram bins (all unique INT trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique INT trigrams occurring at least twice). |
Parson’s Code Trigrams¶
Description: Histogram features for Parson’s Code (PARSON) trigrams: all unique PARSON trigrams and all PARSON trigrams occurring at least twice.
Feature Definition File Name:
PARSON_3GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [-1:+1] |
Var |
Histogram bins (all unique PARSON trigrams). |
|
Integer |
Var |
Histogram bin count (all unique PARSON trigrams). |
|
Array of Integer [-1:+1] |
Var |
Histogram bins (all unique PARSON trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique PARSON trigrams occurring at least twice). |
Semitone Interval Bigrams¶
Description: Histogram features for semitone intervals (INT) bigrams: all unique INT bigrams and all INT bigrams occurring at least twice.
Feature Definition File Name:
INT_2GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer |
Var |
Histogram bins (all unique INT bigrams). |
|
Integer |
Var |
Histogram bin counts (all unique INT bigrams). |
|
Array of Integer |
Var |
Histogram bins (all unique INT bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin counts (all unique INT bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Interval Class Trigrams¶
Description: Histogram features for interval class (IC) trigrams: all unique IC trigrams and all IC trigrams occurring at least twice.
Feature Definition File Name:
INTCLASS_3GRAMS
Category:
Sequence/Interval
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer |
Var |
Histogram bins (all unique IC trigrams). |
|
Integer |
Var |
Histogram bin counts (all unique IC trigrams). |
|
Array of Integer |
Var |
Histogram bins (all unique IC trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique IC trigrams occurring at least twice). |
Sequence/Pitch¶
Pitch Class Bigrams¶
Description: Histogram features for pitch class (PC) bigrams: all unique PC bigrams and all PC bigrams occurring at least twice.
Feature Definition File Name:
PC_2GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique PC bigrams). |
|
Integer |
Var |
Histogram bin count (all unique PC bigrams). |
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique PC bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique PC bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Chordal Diatonic Pitch Class Trigrams¶
Description: Histogram features for Chordal Diatonic Pitch Class (CDPC) trigrams: all unique CDPC trigrams and all CDPC trigrams occurring at least twice.
Feature Definition File Name:
CDPC_3GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
List of Strings |
Var |
Histogram bins (all unique CDPC 3-grams). |
|
Integer |
Var |
Histogram bin count (all unique CDPC 3-grams). |
|
List of String |
Var |
Histogram bins (all unique CDPC 3-grams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique CDPC 3-grams occurring at least twice). |
Chordal Diatonic Pitch Class Bigrams¶
Description: Histogram features for Chordal Diatonic Pitch Class (CDPC) bigrams: all unique CDPC bigrams and all CDPC bigrams occurring at least twice.
Feature Definition File Name:
CDPC_2GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
Var |
Histogram bins (all unique CDPC bigrams). |
|
String |
Var |
Histogram bin counts (all unique CDPC bigrams). |
|
String |
Var |
Histogram bins (all unique CDPC bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin counts (all unique CDPC bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Pitch Class Trigrams¶
Description: Histogram features for pitch class (PC) trigrams: all unique PC trigrams and all PC trigrams occurring at least twice.
Feature Definition File Name:
PC_3GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique PC trigrams). |
|
Integer |
Var |
Histogram bin count (all unique PC trigrams). |
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique PC trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique PC trigrams occurring at least twice). |
Tonal Pitch Class Bigrams¶
Description: Histogram features for raw MIDI pitch (TPC) bigrams: all unique TPC bigrams and all TPC bigrams occurring at least twice.
Feature Definition File Name:
TPC_2GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique TPC bigrams). |
|
Integer |
Var |
Histogram bin count (all unique TPC bigrams). |
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique TPC bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique TPC bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Pitch Bigrams¶
Description: Histogram features for raw MIDI pitch (PITCH) bigrams: all unique PITCH bigrams and all PITCH bigrams occurring at least twice.
Feature Definition File Name:
PITCH_2GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:127] |
Var |
Histogram bins (all unique PITCH bigrams). |
|
Integer |
Var |
Histogram bin count (all unique PITCH bigrams). |
|
Array of Integer [0:127] |
Var |
Histogram bins (all unique PITCH bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique PITCH bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Pitch Pattern Lengths and Distances¶
Description: Average length and occurrence distance between exact pitch patterns. Uses the edit distance based self-similarity matrix alogorithm.
Feature Definition File Name:
PITCH_PATTERN_DIST_LEN_FEATURES
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean length of exact pitch patterns. |
|
Real |
1 |
Mean distance between exact pitch patterns. |
Pitch Trigrams¶
Description: Histogram features for raw MIDI pitch (PITCH) trigrams: all unique PITCH trigrams and all PITCH trigrams occurring at least twice.
Feature Definition File Name:
PITCH_3GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:127] |
Var |
Histogram bins (all unique PITCH trigrams). |
|
Integer |
Var |
Histogram bin count (all unique PITCH trigrams). |
|
Array of Integer [0:127] |
Var |
Histogram bins (all unique PITCH trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique PITCH trigrams occurring at least twice). |
Chordal Pitch Class Bigrams¶
Description: Histogram features for Chordal Pitch Class (CPC) bigrams: all unique CPC bigrams and all CPC bigrams occurring at least twice.
Feature Definition File Name:
CPC_2GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
Var |
Histogram bins (all unique CPC bigrams). |
|
String |
Var |
Histogram bin counts (all unique CPC bigrams). |
|
String |
Var |
Histogram bins (all unique CPC bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin counts (all unique CPC bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Tonal Pitch Class Trigrams¶
Description: Histogram features for tonal pitch class (TPC) trigrams: all unique TPC trigrams and all TPC trigrams occurring at least twice.
Feature Definition File Name:
TPC_3GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique TPC trigrams). |
|
Integer |
Var |
Histogram bin count (all unique TPC trigrams). |
|
Array of Integer [0:11] |
Var |
Histogram bins (all unique TPC trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique TPC trigrams occurring at least twice). |
Chordal Pitch Class Trigrams¶
Description: Histogram features for Chordal Pitch Class (CPC) trigrams: all unique CPC trigrams and all CPC trigrams occurring at least twice.
Feature Definition File Name:
CPC_3GRAMS
Category:
Sequence/Pitch
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
List of Strings |
Var |
Histogram bins (all unique CPC 3-grams). |
|
Integer |
Var |
Histogram bin count (all unique CPC 3-grams). |
|
List of String |
Var |
Histogram bins (all unique CPC 3-grams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique CPC 3-grams occurring at least twice). |
Sequence/Rhythm¶
Duration Class Bigrams¶
Description: Histogram features for (absolute) duration classes (DURCLASS) bigrams: all unique DURCLASS bigrams and all DURCLASS bigrams occurring at least twice.
Feature Definition File Name:
DURCLASS_2GRAMS
Category:
Sequence/Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
Var |
Histogram bins (all unique DURCLASS bigrams). |
|
Integer |
Var |
Histogram bin count (all unique DURCLASS bigrams). |
|
String |
Var |
Histogram bins (all unique DURCLASS bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique DURCLASS bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Duration Class Trigrams¶
Description: Histogram features for (absolute) duration classes (DURCLASS) trigrams: all unique DURCLASS trigrams and all DURCLASS trigrams occurring at least twice. Reference time-span is 0.5s.
Feature Definition File Name:
DURCLASS_3GRAMS
Category:
Sequence/Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
Var |
Histogram bins (all unique DURCLASS trigrams). |
|
Integer |
Var |
Histogram bin count (all unique DURCLASS trigrams). |
|
String |
Var |
Histogram bins (all unique DURCLASS trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique DURCLASS trigrams occurring at least twice). |
Duration Class Run-lengths¶
Description: Mean run-length of duration classes (very short: -2, short: -1, medium: 0, long: 1, very long: 2). Reference time-span is either local beat duration (“relative mode”) or 0.5 sec (“absolute mode”).
Feature Definition File Name:
DURCLASS_RUNLENGTH
Category:
Sequence/Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean run-length averaged over all classes. |
|
Real |
1 |
Mean run-length for very short events. |
|
Real |
1 |
Mean run-length for short events. |
|
Real |
1 |
Mean run-length for medium long events. |
|
Real |
1 |
Mean run-length for long events. |
|
Real |
1 |
Mean run-length for very long events. |
|
Integer |
Var |
Raw vector of run-lengths for all classes. |
Inter-onset Interval Class Bigrams¶
Description: Histogram features for rel. IOI classes (IOICLASS) bigrams: all unique IOICLASS bigrams and all IOICLASS bigrams occurring at least twice.
Feature Definition File Name:
IOICLASS_2GRAMS
Category:
Sequence/Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
Var |
Histogram bins (all unique IOICLASS bigrams). |
|
Integer |
Var |
Histogram bin count (all unique IOICLASS bigrams). |
|
String |
Var |
Histogram bins (all unique IOICLASS bigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique IOICLASS bigrams occurring at least twice). |
|
Real |
1 |
Unnormalized Entropy of bigram distribution in bits. |
|
Real [0,1] |
1 |
Normalized Entropy of bigram distribution. |
Inter-onset Interval Class Run-lengths¶
Description: Mean run-length of inter-onset interval classes (very short: -2, short: -1, medium: 0, long: 1, very long: 2). Reference time-span is either local beat duration (“relative mode”) or 0.5 sec (“absolute mode”).
Feature Definition File Name:
IOICLASS_RUNLENGTH
Category:
Sequence/Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
1 |
Mean run-length averaged over all classes (absolute mode). |
|
Real |
1 |
Mean run-length for very short events (absolute mode). |
|
Real |
1 |
Mean run-length for short events (absolute mode). |
|
Real |
1 |
Mean run-length for medium long events (absolute mode). |
|
Real |
1 |
Mean run-length for long events (absolute mode). |
|
Real |
1 |
Mean run-length for very long events (absolute mode). |
|
Integer |
Var |
Raw vector of run-lengths for all classes (absolute mode). |
|
Real |
1 |
Mean run-length averaged over all classes (relative mode). |
|
Real |
1 |
Mean run-length for very short events (relative mode). |
|
Real |
1 |
Mean run-length for short events (relative mode). |
|
Real |
1 |
Mean run-length for medium long events (relative mode). |
|
Real |
1 |
Mean run-length for long events (relative mode). |
|
Real |
1 |
Mean run-length for very long events (relative mode). |
|
Integer |
Var |
Raw vector of run-lengths for all classes (relative mode). |
Inter-onset Interval Class Trigrams¶
Description: Histogram features for (absolute) IOI classes (IOICLASS) trigrams: all unique IOICLASS trigrams and all IOICLASS trigrams occurring at least twice. Reference time-span is 0.5s.
Feature Definition File Name:
IOICLASS_3GRAMS
Category:
Sequence/Rhythm
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
Var |
Histogram bins (all unique IOICLASS trigrams). |
|
Integer |
Var |
Histogram bin count (all unique IOICLASS trigrams). |
|
String |
Var |
Histogram bins (all unique IOICLASS trigrams occurring at least twice). |
|
Integer |
Var |
Histogram bin count (all unique IOICLASS trigrams occurring at least twice). |
Structure¶
Form Structure of Phrases¶
Description: Calculates form strings based on self-similarity between phrases using edit distance based on interval and duration classes.
Feature Definition File Name:
GENERAL_PHRASE_FORM
Category:
Structure
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
1 |
Phrase form structure based on self-similarity matrix of interval edit distances. |
|
String |
1 |
Phrase form structure based on self-similarity matrix of duration class edit distances. |
Self-similarity Features¶
Description: Features derived from self similarity-matrix between phrases. These are basically statistical descriptors (max, median, stddev, range, entropy) either of the set of adjacent or non-adjacent cells in the self-similarity matrix. For instance, a high median of adjacent cells could indicate developing variation, the same value for non-adjacent cells hints at long-range correlations, i.e. form thinking.
Feature Definition File Name:
GENERAL_SELF_SIM_MATRIX_PHRASES_FEATURES
Category:
Structure
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0,1] |
1 |
Maximum of adjacent similarity values. |
|
Real [0,1] |
1 |
Median of adjacent similarity values. |
|
Real [0,1] |
1 |
Standard deviaation of adjacent similarity values. |
|
Real [0,1] |
1 |
Range of adjacent similarity values. |
|
Real [0,1] |
1 |
Entropy of adjacent similarity values. |
|
Real [0,1] |
1 |
Median of non-adjacent similarity values. |
|
Real [0,1] |
1 |
Standard deviation of non-adjacent similarity values. |
|
Real [0,1] |
1 |
Range of non-adjacent similarity values. |
|
Real [0,1] |
1 |
Maximum of adjacent similarity values. |
Self-similarity Matrix of Phrases¶
Description: Similarity between phrases using edit distance based similarity on pitch values.
Feature Definition File Name:
GENERAL_SELF_SIM_MATRIX_PHRASES
Category:
Structure
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real [0,1] |
Interval-based self-similarity matrix of phrases. |
|
|
Real [0,1] |
Thresholded interval-based self-similarity matrix of phrases. |
Tone Formation¶
Loudness¶
Description: Loudness (resp. intensity) related values for single tone events have been extracted from jazz solos by means of sophisticated MIR techniques and are stored in the WJazzD SQLITE database. Currently, five different values are available: Maximum, median and standard deviation of loudness, relative position of loudness peak and temporal centroid. Might not be avaibable for every solo.
Feature Definition File Name:
GENERAL_LOUDNESS
Category:
Tone Formation
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Float [-120, 0] |
N |
Maximum loudness values per tone. |
|
Real [-120, 0] |
N |
Median loudness values per tone. |
|
Real |
N |
Standard deviation of loudness per tone. |
|
Real [0, 1] |
N |
Relative position of maximum loudness value, duration of tone equals 1. |
|
Real [0, 1] |
N |
Normed temporal centroid of loudness curve per tone, duration of tone equals 1. |
|
Real |
N |
Ratio of solo loudness (median) to backing track loudness (median). |
Articulation¶
Description: Articulation of notes as measured by the ratio of duration and inter-onset interval. (Be careful at phrase endings, though). The higher the ratio, the more legato, the lower the ratio, the more staccato.
Feature Definition File Name:
GENERAL_ARTICULATION
Category:
Tone Formation
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
Real |
N-1 |
Raw interonset-interval values. |
|
Real |
N |
Raw durations. |
|
Real |
N-1 |
Ratios of durations to interonset-intervals. |
|
Real |
1 |
Mean of articulation distribution. |
|
Real |
1 |
Median of articulation distribution. |
|
Real |
1 |
Variance of of articulation distribution. |
|
Real |
1 |
Standard deviation of articulation distribution. |
|
Real |
1 |
Minimum of of articulation distribution. |
|
Real |
1 |
Maximum of of articulation distribution. |
|
Real |
1 |
Range of of articulation distribution. |
Modulation¶
Description: F0-modulation of notes. Possible annoated values are
fall-off
,slide
,vibrato
,bend
,straight
, or an empty string, which might indicate that either no modulation is present or that no articulation was annotated. There are furthermore three automatically extracted value available: f0_mod_range_cents (Modulations in cents), f0_mod_freq_hz (Modulation frequency in Hz), f0_median_dev (Median devation from nominal 12TET pitch in cents).Feature Definition File Name:
GENERAL_F0MODULATION
Category:
Tone Formation
Included Features:
Label |
Type |
Length |
Description |
---|---|---|---|
|
String |
N |
Annonated modulation per tone. |
|
String |
N |
Annonated modulation per tone (short label). |
|
Real |
N |
Extracted modulation range (cents). |
|
Real |
N |
Extracted modulation range (Hz). |
|
Real |
N |
Extracted median deviation from nominal 12TET pitch (cents). |