TABLE 3.
Audio features extracted from Timbre Toolbox.
| Audio features from Timbre Toolbox | ||
| Representation | Feature | Description |
| STFT | Spectral centroid | Center of gravity of the spectrum |
| STFT | Spectral spread | Standard deviation of the spectrum around the mean |
| STFT | Spectral skewness | Asymmetry of the spectrum around the mean |
| STFT | Spectral kurtosis | Flatness of the spectrum around the mean |
| STFT | Spectral flatness | Ratio of the geometric and arithmetic means of the spectrum |
| STFT | Spectral crest | Ratio of the spectral maximum to the arithmetic spectral mean |
| STFT | Spectral slope | Linear regression over the spectral amplitude values |
| STFT | Spectral decrease | Average of slopes between F0 and 2nd to kth harmonic |
| STFT | Spectral roll-off | Frequency below which 95% of the signal energy is contained |
| STFT | Spectral variation | A measure of variability of the spectrum over time: correlation between spectra in successive time frames |
| STFT | Spectral flux | A measure of variability of the spectrum over time: Euclidean distance between spectra in successive time frames |
| HARM | F0 | Fundamental frequency of a periodic sound |
| HARM | Harmonic spectral deviation | Deviation of the amplitudes of the partials from a smoothed spectral envelope |
| HARM | Tristimulus 1 | Ratio of energy of the 1st harmonic to total energy |
| HARM | Tristimulus 2 | Ratio of energy of the 2nd, 3rd, and 4th harmonics to total energy |
| HARM | Tristimulus 3 | Ratio of energy of remaining harmonics (above 4th) to total energy |
| HARM | Harmonic odd-to-even ratio | Ratio of energy of odd harmonics to even harmonics |
| HARM | Inharmonicity | Degree to which frequencies of overtones depart from multiples of the fundamental frequency |
| HARM | Harmonic energy | Energy of the signal explained by stable partials |
| HARM | Noise energy | Energy of the signal not explained by stable partials |
| HARM | Noisiness | Ratio of noise energy to total energy |
| HARM | Harmonic-to-noise ratio | Ratio between periodic and non-periodic components of a signal |
| TEE | Attack time | Duration of the attack portion of the sound |
| TEE | Log attack time | Logarithm of the duration of the attack portion of the sound |
| TEE | Attack slope | Rate of change of energy over time in the attack portion |
| TEE | Decrease slope | Measure of the rate of decrease of the signal energy |
| TEE | Temporal centroid | Center of gravity of the energy envelope |
| TEE | Effective duration | Time during which energy envelope is above 40% (intended to reflect perceived duration) |
| TEE | Frequency of energy modulation | Frequency of the modulation of energy over the sustained portion of the sound as represented using a sinusoidal component |
| TEE | Amplitude of energy modulation | Amplitude of the modulation of energy over the sustained portion of the sound as represented using a sinusoidal component |
STFT, short-time Fourier transform; HARM, harmonic; TEE, temporal energy envelope. For further detail on how features are computed, see Kazazis et al. (2021).