jacquesmartell / SpectralFeatures

Spectral features extraction for audio analysis.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Spectral Features

Statistical and spectral features extraction for audio analysis.


Statistical features

Obtained using the time domain.

Feature Description
Energy Sum of the squared amplitude of the signal.
Power Sum of the squared amplitude over one period of the signal.
Root-mean-square Sum of the absolute value of the amplitude over one period of the signal.
Zero–crossing rate Average rate of change of the signal’s sign.

Spectral shape features

Obtained using the frequency domain.

Feature Description Formula
Spectral centroid Energetic center of the signal.
Spectral spread Deviation of the energy of the signal with respect to the centroid.
Spectral skewness Lateral localization of the energy of the signal.
Spectral kurtosis Weigthed presence of the outliers in the signal’s tails.
Spectral rolloff Estimation of the cutoff frequency of the signal.
Spectral decrease Average drop of the signal’s energy.
Spectral slope Degree of the steepness of the signal.
Spectral peak Frequency with highest energy concentration within the signal.
Spectral crest Ratio of the spectral peak to the signal’s mean.
Spectral entropy White noise percentage of the signal.
Spectral flatness Ratio of the signal’s geometric mean to its arithmetic mean.

About

Spectral features extraction for audio analysis.


Languages

Language:Python 54.7%Language:MATLAB 23.8%Language:C++ 21.5%