-
-
Notifications
You must be signed in to change notification settings - Fork 34
Features
Mohammed Boujemaoui Boulaghmoudi edited this page Oct 6, 2018
·
1 revision
Feature engineering is the process of using domain knowledge of the data to create features that make machine learning algorithms work. This section introduce a list of available features in the MIR (Music Information Retrieval) module.
Those are some of the possible applications of those sets of features:
- Genre classification
- Mood classification
- Music Recommendation
- Artist identification
- Artist similarity
- Cover song detection
- Rhythm and beat detection
- Score following
- Chord detection
- Organization of music
- Audio Fingerprinting
- Audio segmentation
- Instrument detection
- Automatic source separation
- Onset detection
- Optical music recognition
- Melody transcription
Information related with the original sound file.
| Descriptor Name | Frame | Size | Tag | State |
|---|---|---|---|---|
| Singer | Y | 1 | Singer |
X |
| Year | Y | 1 | Year |
X |
| Album | Y | 1 | Album |
X |
| Tag | Y | 1 | Tag |
X |
Information related with the statistical properties of the signal.
| Descriptor Name | Frame | Size | Tag | State |
|---|---|---|---|---|
| Arithmetic Mean | Y | 1 | ArithmeticMean |
X |
| Geometric Mean | Y | 1 | GeometricMean |
X |
| Harmonic Mean | Y | 1 | HarmonicMean |
X |
| Generalized Mean | Y | 1 | GeneralizedMean |
X |
| Centroid | Y | 1 | Centroid |
X |
| Variance | Y | 1 | Variance |
X |
| Standard Deviation | Y | 1 | StandardDeviation |
X |
| Skewness | Y | 1 | Skewness |
X |
| Kurtosis | Y | 1 | Kurtosis |
X |
| Generalized Central Moments | Y | 1 | CentralMoments |
X |
| Quantiles | Y | 1 | Quantiles |
X |
Features computed from the waveform of the signal energy (envelop), including the ones referring to various energy content of the signal.
| Descriptor Name | Frame | Size | Tag | State |
|---|---|---|---|---|
| Log Attack Time | N | 1 | LogAttackTime |
X |
| Temporal Increase | N | 1 | TemporalIncrease |
X |
| Temporal Decrease | N | 1 | TemporalDecrease |
X |
| Temporal Centroid | N | 1 | TemporalCentroid |
X |
| Duration | Y | 1 | Duration |
X |
| Effective Duration | Y | 1 | EffectiveDuration |
X |
| Energy | Y | 1 | Energy |
X |
| Harmonic Energy | Y | 1 | HarmonicEnergy |
X |
| Noise Energy | Y | 1 | Noise Energy |
X |
| Auto-Correlation | Y | 12 | AutoCorrelation |
X |
| Zero-Crossing Rate | Y | 1 | ZeroCrosssingRate |
X |
Features strongly related to the harmonic properties of the signal.
| Descriptor Name | Frame | Size | Tag | State |
|---|---|---|---|---|
| Tonalness | Y | 1 | Tonalness |
X |
| Key | Y | 1 | Key |
X |
| Onsets | Y | 1 | Onsets |
X |
| Tempo | Y | 1 | Tempo |
X |
| Meter | Y | 1 | Meter |
X |
| Rhytm | Y | 1 | Rhytm |
X |
| Timing | Y | 1 | Timing |
X |
| Pitch Hz | Y | 1 | PitchHz |
X |
| Pitch Midi | Y | 1 | PitchMidi |
X |
| Note Hz | Y | 1 | NoteHz |
X |
| Note Midi | Y | 1 | NoteMidi |
X |
| Beat | Y | 1 | Beat |
X |
| Mood | Y | 1 | Mood |
X |
| Beat | Y | 1 | Beat |
X |
| Beat | Y | 1 | Beat |
X |
Features computed from the Short Time Fourier Transform (STFT) of the signal.
| Descriptor Name | Frame | Size | Tag | State |
|---|---|---|---|---|
| Spectral Centroid | Y | 6 | SpectrumCentroid |
X |
| Spectral Spread | Y | 6 | SpectrumSpread |
X |
| Spectral Skewness | Y | 6 | SpectrumSkewness |
X |
| Spectral Kurtosis | Y | 6 | SpectrumKurtosis |
X |
| Spectral Slope | Y | 6 | SpectrumSlope |
X |
| Spectral Decrease | Y | 1 | SpectrumDecrease |
X |
| Spectral Roll-Off | Y | 1 | SpectrumRollOff |
X |
| Spectral Variation | Y | 3 | SpectrumVariation |
X |
| Spectral Flatness | Y | 4 | SpectrumFlatness |
X |
| Spectral Flux | Y | 1 | SpectrumFlux |
X |
| Spectral Crest | Y | 4 | SpectrumCrest |
X |
| MFCC | Y | 12 | MFCC |
X |
| Delta MFCC | Y | 12 | DeltaMFCC |
X |
| Delta Delta MFCC | Y | 12 | DeltaDeltaMFCC |
X |
| LPC | Y | 12 | LPC |
X |
| LPCC | Y | 12 | LPCC |
X |
Features computed from the Sinusoidal Harmonic modeling of the signal.
| Descriptor Name | Frame | Size | Tag | State |
|---|---|---|---|---|
| Fundamental Frequency | Y | 1 | FundamentalFrequency |
X |
| Noisiness | Y | 1 | Harmonicity |
X |
| In-Harmonicity | Y | 1 | Inharmonicity |
X |
| Harmonic Spectral Deviation | Y | 3 | HarmonicSpectralDeviation |
X |
| Odd-to-Even Harmonic Ratio | Y | 3 | HarmonicSpectralOERatio |
X |
| Harmonic Tristimulus | Y | 9 | HarmonicSpectralTristimulus |
X |
| Harmonic Spectral Centroid | Y | 6 | HarmonicSpectralCentroid |
X |
| Harmonic Spectral Spread | Y | 6 | HarmonicSpectralSpread |
X |
| Harmonic Spectral Skewness | Y | 6 | HarmonicSpectralSkewness |
X |
| Harmonic Spectral Kurtosis | Y | 6 | HarmonicSpectralKurtosis |
X |
| Harmonic Spectral Slope | Y | 6 | HarmonicSpectralSlope |
X |
| Harmonic Spectral Decrease | Y | 1 | HarmonicSpectralDecrease |
X |
| Harmonic Spectral Roll-Off | Y | 1 | HarmonicSpectralRollOff |
X |
| Harmonic Spectral Variation | Y | 3 | HarmonicSpectralVariation |
X |
Features computed using a model of the human earring process.
| Descriptor Name | Frame | Size | Tag | State |
|---|---|---|---|---|
| Loudness | Y | 1 | Loudness |
X |
| Relative Specific Loudness | Y | 24 | RelativeSpecificLoudness |
X |
| Sharpness | Y | 1 | Sharpness |
X |
| Spread | Y | 1 | Spread |
X |
| Perceptual Spectral Centroid | Y | 6 | FilterbankCentroid |
X |
| Perceptual Spectral Spread | Y | 6 | FilterbankSpread |
X |
| Perceptual Spectral Skewness | Y | 6 | FilterbandSkewness |
X |
| Perceptual Spectral Kurtosis | Y | 6 | FilterbankKurtosis |
X |
| Perceptual Spectral Slope | Y | 6 | FilterbankSlope |
X |
| Perceptual Spectral Decrease | Y | 1 | FilterbankDecrease |
X |
| Perceptual Spectral Roll-Off | Y | 1 | FilterbankRollOff |
X |
| Perceptual Spectral Variation | Y | 3 | FilterbankVariation |
X |
| Odd to Even Band Ratio | Y | 3 | FilterbankOERatio |
X |
| Band Spectral Deviation | Y | 3 | FilterbankDeviation |
X |
| Band Tristimulus | Y | 9 | FilterbankTristimulus |
X |
easy DSP - A cross-platform Digital Signal Processing library written in modern C++.