de.sciss.strugatzki.FeatureSegmentation
The size of the sliding window over which the features are correlated.
The size of the sliding window over which the features are correlated. That is, for a length of 1.0 second (given in sample frames, hence 44100 for a sample rate of 44100 Hz), at any given point in time, 0.5 seconds left of that point are correlated with 0.5 seconds right of that point. Breaking points are those where correlation is minimised.
The database folder is merely used to retrieve the normalization file,
given that normalize
is true
.
The XML file holding the extractor parameters corresponding to the audio input file.
The XML file holding the extractor parameters corresponding to the audio input file. The audio input file's feature vector output file is determined from this meta file.
Minimum spacing between breaks
Whether to apply normalization to the features (recommended)
Maximum number of breaks to report
An option which restricts segmentation to a given span within the input file.
An option which restricts segmentation to a given span within the
input file. That is, only breaking points within this span are
reported. If None
, the whole file is considered.
The balance between the feature of loudness curve and spectral composition (MFCC).
The balance between the feature of loudness curve and spectral composition (MFCC). A value of 0.0 means the segmentation is only performed by considering the spectral features, and a value of 1.0 means the segmentation is taking only the loudness into consideration. Values in between give a measure that takes both features into account with the given priorities.
All durations, spans and spacings are given in sample frames with respect to the sample rate of the audio input file.