maad.features.spectral_moments

maad.features.spectral_moments(X, axis=None)[source]

Computes the first 4th moments of an amplitude spectrum (1d) or spectrogram (2d), mean, variance, skewness, kurtosis.

Parameters:

Xndarray of floats: Amplitude spectrum (1d) or spectrogram (2d).
axisinterger, optional, default is None: if spectrogram (2d), select the axis to estimate the moments.

Returns:

meanfloat: mean of the audio
varfloat: variance of the audio
skewfloat: skewness of the audio
kurtfloat: kurtosis of the audio

Examples

>>> from maad import sound, features
>>> s, fs = sound.load('../data/spinetail.wav')
>>> Sxx_power,_,_,_ = sound.spectrogram(s, fs)

Compute spectral moments on the mean spectrum

>>> import numpy as np
>>> S_power = sound.avg_power_spectro(Sxx_power)
>>> sm, sv, ss, sk = features.spectral_moments (S_power)
>>> print('mean: %2.8f / var: %2.10f / skewness: %2.2f / kurtosis: %2.2f' % (sm, sv, ss, sk))
mean: 0.00000228 / var: 0.0000000001 / skewness: 5.84 / kurtosis: 40.49

Compute spectral moments of the spectrogram along the time axis

>>> from maad.features import spectral_moments
>>> sm_per_bin, sv_per_bin, ss_per_bin, sk_per_bin = spectral_moments(Sxx_power, axis=1)
>>> print('Length of sk_per_bin is : %2.0f' % len(sk_per_bin))
Length of sk_per_bin is : 512