maad.features.spectral_entropy

maad.features.spectral_entropy(Sxx, fn, flim=None, display=False)[source]

Compute different entropies based on the average spectrum, its variance, and its maxima [1] [2]

Parameters:

Sxxndarray of floats: Spectrogram (2d). It is recommended to work with PSD to be consistent with energy conservation
fn1d ndarray of floats: frequency vector
flimtupple (fmin, fmax), optional, default is None: Frequency band used to compute the spectral entropy. For instance, one may want to compute the spectral entropy for the biophony bandwidth
displayboolean, optional, default is False: Display the different spectra (mean, variance, covariance, max…)

Returns:

EASscalar: Entropy of Average Spectrum
ECUscalar: Entropy of spectral variance (along the time axis for each frequency)
ECVscalar: Entropy of Coefficient of Variation (along the time axis for each frequency)
EPSscalar: Entropy of spectral maxima (peaks)
EPS_KURTscalar: Kurtosis of spectral maxima
EPS_SKEWscalar: Skewness of spectral maxima

References

[1]

TOWSEY, Michael W. The calculation of acoustic indices derived from long-duration recordings of the natural environment. 2017. https://eprints.qut.edu.au/110634/1/QUTePrints110634_TechReport_Towsey2017August_AcousticIndices%20v3.pdf

[2]

QUT : https://github.com/QutEcoacoustics/audio-analysis. Michael Towsey, Anthony Truskinger, Mark Cottman-Fields, & Paul Roe. (2018, March 5). Ecoacoustics Audio Analysis Software v18.03.0.41 (Version v18.03.0.41). Zenodo. http://doi.org/10.5281/zenodo.1188744

Examples

>>> import maad
>>> s, fs = maad.sound.load('../data/cold_forest_daylight.wav')
>>> Sxx_power, tn, fn, _ = maad.sound.spectrogram (s, fs)  
>>> EAS, ECU, ECV, EPS, EPS_KURT, EPS_SKEW = maad.features.spectral_entropy(Sxx_power, fn, flim=(2000,10000)) 
>>> print('EAS: %2.2f / ECU: %2.2f / ECV: %2.2f / EPS: %2.2f / EPS_KURT: %2.2f / EPS_SKEW: %2.2f' % (EAS, ECU, ECV, EPS, EPS_KURT, EPS_SKEW)) 
EAS: 0.27 / ECU: 0.49 / ECV: 0.24 / EPS: 0.25 / EPS_KURT: 17.58 / EPS_SKEW: 3.55