pyrem.feature_families module¶

The goal of this submodule is to provide a flexible interface to compute arbitrary features on each channel and epoch (temporal slices) of a multivariate time series (Polygraph). Features are grouped in families of several features (e.g. Power Features may contain mean power, variance of power, ...). Feature factory computes features for arbitrary feature families and group them in a data.frame

class pyrem.feature_families.AbsoluteFeatures[source]¶

Bases: pyrem.feature_families.SignalFeatureBase

prefix = 'abs'¶

class pyrem.feature_families.AnnotationFeatureBase[source]¶: Bases: pyrem.feature_families.FeatureFamilyBase

class pyrem.feature_families.EntropyFeatures[source]¶

Bases: pyrem.feature_families.SignalFeatureBase

prefix = 'entropy'¶

class pyrem.feature_families.FeatureFamilyBase[source]¶

Bases: object

A feature family object is a process returning a vector of features upon analysis of some data. Features are returned as a pandas DataFrame object, with column names for features. Each feature name is prefixed by the name of the feature family. This is an abstract class designed to be derived by:

Defining a prefix attribute. It will add the name of the family to the name of the features.
Overriding the _make_feature_vec method. It should return a dictionary of scalars, each being a feature.

make_vector(signal)[source]¶

Compute one vector of features from polygraph.