selfeeg.augmentation

This is the data augmentation module of the selfEEG library. It is divided in two submodules:

• Functional: a collection of data augmentations compatible with numpy arrays and torch tensors moved to both CPU or GPU devices.

• Compose: a collection of classes designed to combine data augmentations in complex patterns.

augmentation.compose module

Classes

CircularAug	Single Augmenter called sequentially from a list, following a circular order.
DynamicSingleAug	Single augmentation with randomly chosen arguments.
RandomAug	Random augmentation chosen from a given set.
SequentialAug	Multiple augmentations applied sequentially.
StaticSingleAug	Single static augmentation with preset arguments.

augmentation.functional module

Functions

add_band_noise	adds random noise filtered at specific bandwidths.
add_eeg_artifact	add common EEG artifacts to the ArrayLike object.
add_gaussian_noise	adds gaussian noise with the desired standard deviation.
add_noise_SNR	add noise to the input ArrayLike object such that its SNR (Signal to Noise Ratio) will be the one desired.
change_ref	changes the reference of all EEG record in the ArrayLike object.
channel_dropout	puts to 0 a given (or random) amount of channels of the ArrayLike object.
crop_and_resize	crops some segments of the ArrayLike object.
filter_bandpass	applies a bandpass filter on the last dimension of the ArrayLike.
filter_bandstop	applies a bandstop filter on the last dimension of the ArrayLike object.
filter_highpass	applies an highpass filter on the last dimension of the ArrayLike object.
filter_lowpass	applies a lowpass filter on the last dimension of the ArrayLike.
flip_horizontal	flips the elements of the ArrayLike object along its last dimension.
flip_vertical	changes the sign of all the elements of the input.
get_channel_map_and_networks	returns the channel_map and the chan_net arguments for permute_channels.
get_filter_coeff	returns the coefficients a and b necessary to run filtering augmentations.
identity	returns the same array or tensor it was given.
masking	puts to zero random portions of the ArrayLike object.
moving_avg	applies a moving average filter to the input ArrayLike object.
permutation_signal	permutes some portions of the ArrayLike object along its last dimension.
permute_channels	permutes the ArrayLike object along the EEG channel dimension.
phase_swap	Apply the phase swap data augmentation to the ArrayLike object.
random_FT_phase	randomizes the phase of all signals in the input ArrayLike object.
random_slope_scale	randomly scales the first derivative of x.
scaling	rescales the ArrayLike object by a given amplitude.
shift_frequency	shifts the frequency components of the signals included in the ArrayLike object.
shift_horizontal	Shifts temporally the elements of the ArrayLike object.
shift_vertical	adds a scalar value to the ArrayLike object.
warp_signal	stretches and squeezes portions of the ArrayLike object.