Transformations¶

Audio signal transformations in Sigment are represented by classes that can be used to apply a specific type of transformation to the audio data.

Some example transformation classes include GaussianWhiteNoise and TimeStretch. A full list of available transformations and their details and parameters can be found below.

Each of these transformation classes are a subclass of the more generic Transform base class, which provides a basic interface that can also be used to write custom transformations.

Sigment offers a familiar interface for transformations, taking inspiration from popular augmentation libraries such as imgaug, nlpaug, albumentations and audiomentations.

Section contents

Available transformations
Using transformations

Available transformations ¶

Identity ¶

class sigment.transforms.Identity[source]¶

Applies an identity transformation to a signal.

Notes

A sampling rate is not required when applying this transformation.

Additive Gaussian White Noise ¶

class sigment.transforms.GaussianWhiteNoise(scale, p=1.0, random_state=None)[source]¶

Applies additive Gaussian white noise to the signal.

Parameters

scale: float [scale > 0] or tuple: Amount to scale the value sampled from the standard normal distribution.

Essentially the variance $\sigma^2$.

Notes

A sampling rate is not required when applying this transformation.

Time Stretch ¶

class sigment.transforms.TimeStretch(rate, p=1.0, random_state=None)[source]¶

Stretches the duration or speed of the signal without affecting its pitch.

Parameters

rate: float [rate > 0] or tuple

Stretch rate.

If rate < 1, the signal is slowed down.
If rate > 1, the signal is sped up.

Notes

A sampling rate is not required when applying this transformation.

Pitch Shift ¶

class sigment.transforms.PitchShift(n_steps, p=1.0, random_state=None)[source]¶

Shifts the pitch of the signal without changing its duration or speed.

Parameters

n_steps: float [-12 < n_steps < 12] or tuple: Number of semitones to shift.

Notes

A sampling rate is required when applying this transformation.

Edge Crop ¶

class sigment.transforms.EdgeCrop(side, crop_size, p=1.0, random_state=None)[source]¶

Crops a section from the start or end of the signal.

Parameters

side: {‘start’, ‘end’}: The side of the signal to crop.
crop_size: float [0 < crop_size < 1] or tuple: The fraction of the signal duration to crop from the chosen side.

Notes

A sampling rate is not required when applying this transformation.

Random Crop ¶

class sigment.transforms.RandomCrop(crop_size, n_crops, p=1.0, random_state=None)[source]¶

Randomly crops multiple sections from the signal.

Parameters

crop_size: float [0 < crop_size < 1] or tuple: The fraction of the signal duration to crop.
n_crops: int [n_crops > 0] or tuple: The number of random crops of size crop_size to make.

Notes

Chunking is done according to the algorithm defined at [1].
crop_size $\times$ n_crops must not exceed 1.
A sampling rate is not required when applying this transformation.

References

1: https://stackoverflow.com/a/49944026

Linear Fade In/Out ¶

class sigment.transforms.LinearFade(direction, fade_size, p=1.0, random_state=None)[source]¶

Linearly fades the signal in or out.

Parameters

direction: {‘in’, ‘out’}: The direction to fade the signal.
fade_size: float [0 < fade_size < 0.5] or tuple: The fraction of the signal to fade in the chosen direction.

Notes

A sampling rate is not required when applying this transformation.

Normalize ¶

class sigment.transforms.Normalize(independent=True, p=1.0, random_state=None)[source]¶

Normalizes the signal by dividing each sample by the maximum absolute sample amplitude.

Parameters

independent: bool: Whether or not to normalize each channel independently.

Notes

A sampling rate is not required when applying this transformation.

Pre-Emphasize ¶

class sigment.transforms.PreEmphasize(alpha=0.95, p=1.0, random_state=None)[source]¶

Pre-emphasizes the signal by applying a first-order high-pass filter.

\[\begin{split}x'[t] = \begin{cases} x[t] & \text{if $t=0$} \\ x[t] - \alpha x[t-1] & \text{otherwise} \end{cases}\end{split}\]

Parameters

alpha: float [0 < alpha < 1] or tuple: Pre-emphasis coefficient.

Notes

A sampling rate is not required when applying this transformation.

Extract Loudest Section ¶

class sigment.transforms.ExtractLoudestSection(duration, p=1.0, random_state=None)[source]¶

Extracts the loudest section from the signal using sliding window aggregation over amplitudes.

Parameters

duration: float [0 < duration < 1] or tuple: The duration of the section to extract, as a fraction of the original signal duration.

Notes

See [2] for more details on the implementation.
A sampling rate is not required when applying this transformation.

References

2: https://github.com/petewarden/extract_loudest_section

Median Filter ¶

class sigment.transforms.MedianFilter(window_size, p=1.0, random_state=None)[source]¶

Applies a median filter to the signal.

\[x'[t] = \mathrm{median} \underbrace{\Big[ \ldots, x[t-1], x[t], x[t+1], \ldots \Big]}_\text{window size}\]

Parameters

window_size: int [window_size > 1] or tuple: The size of the window of neighbouring samples.

Notes

A sampling rate is not required when applying this transformation.

Using transformations ¶

Each transformation class comes with a number of methods that can be used to apply the transformation to either a numpy.ndarray or WAV file.

All transformations accept the p and random_state parameters, inherited from the Transform base class described below.

class sigment.transforms.Transform(p, random_state)[source]¶

Base class representing a single transformation or augmentation.

Note

As this is a base class, it should not be directly instantiated.

You can however, use it to create your own transformations, following the implementation of the pre-defined transformations in Sigment.

Parameters

p: float [0 <= p <= 1]: The probability of executing the transformation.
random_state: numpy.RandomState, int, optional: A random state object or seed for reproducible randomness.

__call__(self, X, sr=None)[source]¶

Runs the transformation on a provided input signal.

Parameters

X: numpy.ndarray [shape (T,) or (1xT) for mono, (2xT) for stereo]: The input signal to transform.
sr: int, optional: The sample rate for the input signal.

Note

Not required if the transformation does not rely on sr.

Returns

transformed: numpy.ndarray [shape (T,) for mono, (2xT) for stereo]: The transformed signal.

Note

If a mono signal X of shape (1xT) was used, the output is reshaped to (T,).

Examples

>>> import numpy as np
>>> from sigment.transforms import PitchShift
>>> # Create an example stereo signal.
>>> X = np.array([
>>>     [0.325, 0.53 , 1.393, 1.211],
>>>     [1.21 , 0.834, 1.022, 0.38 ]
>>> ])
>>> # Create the pitch-shifting transformation object.
>>> shift = PitchShift(n_steps=(-1., 1.))
>>> # Run the __call__ method on the transformation object to transform X.
>>> # NOTE: Pitch shifting requires a sample rate when called.
>>> X_shift = shift(X, sr=10)

generate(self, X, n, sr=None)[source]¶

Runs the transformation on a provided input signal, producing multiple augmented copies of the input signal.

Parameters

X: numpy.ndarray [shape (T,) or (1xT) for mono, (2xT) for stereo]: The input signal to transform.
n: int [n > 0]: Number of augmented copies of X to generate.
sr: int, optional: The sample rate for the input signal.

Note

Not required if not using transformations that require a sample rate.

Returns

augmented: List[numpy.ndarray] or numpy.ndarray: The augmented copies (or copy if n=1) of the signal X.

Note

If a mono signal X of shape (1xT) was used, the output is reshaped to (T,).

Examples

>>> import numpy as np
>>> from sigment.transforms import GaussianWhiteNoise
>>> # Create an example stereo signal.
>>> X = np.array([
>>>     [0.325, 0.53 , 1.393, 1.211],
>>>     [1.21 , 0.834, 1.022, 0.38 ]
>>> ])
>>> # Create the Gaussian white noise transformation object.
>>> add_noise = GaussianWhiteNoise(scale=(0.05, 0.15))
>>> # Generate 5 augmented versions of X, using the noise transformation.
>>> Xs_noisy = add_noise.generate(X, n=5)

apply_to_wav(self, source, out=None)¶

Applies the augmentation to the provided input WAV file and writes the resulting signal back to a WAV file.

Parameters

source: str, Path or path-like: Path to the input WAV file.
out: str, Path or path-like: Output WAV path for the augmented signal.

Warning

If out is set to None (which is the default) or the same as source, the input WAV file will be overwritten!

Examples

>>> import numpy as np
>>> from sigment.transforms import Identity
>>> # Create the identity transformation object.
>>> identity = Identity()
>>> # Apply the transformation to the input WAV file and write it to the output file
>>> identity.apply_to_wav('in.wav', 'out.wav')

generate_from_wav(self, source, n=1)¶

Applies the augmentation to the provided input WAV file and returns a numpy.ndarray.

Parameters

source: str, Path or path-like: Path to the input WAV file.
n: int [n > 0]: Number of augmented versions of the source signal to generate.

Returns

augmented: List[numpy.ndarray] or numpy.ndarray: The augmented versions (or version if n=1) of the source signal.

Examples

>>> import numpy as np
>>> from sigment.transforms import LinearFade
>>> # Create the fade-in transformation object.
>>> fade_in = LinearFade(direction='in', fade_size=(0.025, 0.1))
>>> # Generate 5 augmented versions of the signal data from 'signal.wav' as numpy.ndarrays, using the fade-in transformation.
>>> Xs_faded = fade_in.generate_from_wav('signal.wav', n=5)

Quantifiers¶

Quantifiers are used to specify rules for how a sequence of transformations or quantifiers should be applied.

Each quantifier class is a subclass of the more generic Quantifier base class, which provides a basic interface that can also be used to write custom quantifiers, though there is rarely a need for this.

As with transformations, Sigment offers a familiar interface for quantifiers, taking inspiration from popular augmentation libraries such as imgaug and nlpaug.

Section contents

Available quantifiers
Using quantifiers

Available quantifiers ¶

In the below table, the steps argument is of type List[Transform, Quantifier], specifying a sequence of transformations or quantifiers to be applied.

Quantifier	Summary
Quantifier (Base) `Quantifier(steps, **kwargs)`	A base class representing a single quantifier.
	Main parameters • None
	Notes: As this is a base class, it should not be initialized.

Pipeline `Pipeline(steps, **kwargs)`	Sequentially executes each transformation or quantifier step.
	Main parameters • None
	Notes: None

Sometimes `Sometimes(steps, p, **kwargs)`	Probabilistically applies the provided transformation or quantifier steps.
	Main parameters • p: `0 <= float <= 1` The probability of executing the transformation or quantifier steps.
	Notes: None

Some `SomeOf(steps, n, **kwargs)`	Randomly applies a number of the provided transformation or quantifier steps.
	Main parameters • n: `tuple` or `int > 0` The number of transformation or quantifier steps to apply.
	Notes: The chosen steps will still be applied in the same order they were defined by default.

One `OneOf(steps, **kwargs)`	Randomly applies a single step from the provided transformation or quantifier steps.
	Main parameters • None
	Notes: This is a special case of the `SomeOf` quantifier, with $n=1$.

Using quantifiers ¶

class sigment.quantifiers.Quantifier(steps, [main params, ]random_order=False, random_state=None)¶

Base class representing a single quantifier.

Note

As Quantifier is a base class, it should not be directly instantiated – use one of the quantifier classes listed above.

Parameters

steps (List[Transform, Quantifier]) – A collection of transformation or quantifier steps to apply.
random_order (bool) – Whether or not to randomize the order of execution of steps.
random_state (numpy.RandomState, int or None) – A random state object or seed for reproducible randomness.

__call__(self, X, sr=None)¶

Runs the quantifier steps on a provided input signal.

Parameters

X (numpy.ndarray $(T,)$ or $(1\times T)$ for mono, $(2\times T)$ for stereo) – The input signal to transform.
sr (int $> 0$ or None) – Sample rate.
If the steps of the quantifier do not depend on a sample rate, this should be None (which is the default). See the transformations table to determine whether you need a sample rate or not.

Returns

The transformed signal.

Return type

numpy.ndarray $(T,)$ for mono, $(2\times T)$ for stereo

Example:

import numpy as np
from sigment.quantifiers import SomeOf
from sigment.transforms import GaussianWhiteNoise, PitchShift, EdgeCrop

# Create an example stereo signal.
X = np.array([
    [0.325, 0.53 , 1.393, 1.211],
    [1.21 , 0.834, 1.022, 0.38 ]
])

# Use the SomeOf quantifier to run only 1 to 2 of the transformations.
transform = SomeOf([
    GaussianWhiteNoise(scale=(0.05, 0.15)),
    PitchShift(n_steps=(-1., 1.)),
    EdgeCrop(side='start', crop_size=(0.02, 0.05))
], n=(1, 2))

# Run the __call__ method on the quantifier object to transform X.
# NOTE: Pitch shifting requires a sample rate when called, therefore
#   we must call the quantifier with a specified sample rate parameter.
X_transform = transform(X, sr=10)

generate(self, X, n, sr=None)¶

Runs the quantifier steps on a provided input signal, producing multiple augmented copies of the input signal.

Parameters

X (numpy.ndarray $(T,)$ or $(1\times T)$ for mono, $(2\times T)$ for stereo) – The input signal to transform.
n (int $> 0$) – Number of augmented versions of X to generate.
sr (int $> 0$ or None) – Sample rate.
If the steps of the quantifier do not depend on a sample rate, this should be None (which is the default). See the transformations table to determine whether you need a sample rate or not.

Returns

The augmented versions (or version if n=1) of the signal X.

Return type

List[numpy.ndarray] or numpy.ndarray

Example:

import numpy as np
from sigment.quantifiers import Sometimes, OneOf
from sigment.transforms import LinearFade, GaussianWhiteNoise

# Create an example stereo signal.
X = np.array([
    [0.325, 0.53 , 1.393, 1.211],
    [1.21 , 0.834, 1.022, 0.38 ]
])

# Use the Sometimes and OneOf quantifiers to sometimes (with probability 0.65)
# apply a Gaussian white noise transformation and either a fade-in or fade-out.
transform = Sometimes([
    GaussianWhiteNoise(scale=(0.05, 0.15)),
    OneOf([
        LinearFade(direction='in', fade_size=(0.05, 0.1)),
        LinearFade(direction='out', fade_size=(0.05, 0.1))
    ])
], p=0.65)

# Generate 5 augmented versions of X, using the quantifier object.
Xs_transform = transform.generate(X, n=5)

apply_to_wav(self, source, out=None)¶

Runs the quantifier steps on a provided input WAV file and writes the resulting signal back to a WAV file.

Warning

If out is set to None (which is the default) or the same as source, the input WAV file will be overwritten!

Parameters

source (str, Path or path-like) – Path to the input WAV file.
out (str, Path or path-like) – Output WAV path for the augmented signal.

Example:

from librosa import load
from sigment import *

# Load the stereo WAV audio file
X, sr = load('audio.wav', mono=False)

# Create a complex augmentation pipeline
transform = Pipeline([
    GaussianWhiteNoise(scale=(0.001, 0.0075), p=0.65),
    ExtractLoudestSection(duration=(0.85, 0.95)),
    OneOf([
        RandomCrop(crop_size=(0.01, 0.04), n_crops=(2, 5)),
        SomeOf([
            EdgeCrop('start', crop_size=(0.05, 0.1)),
            EdgeCrop('end', crop_size=(0.05, 0.1))
        ], n=(1, 2))
    ]),
    Sometimes([
        SomeOf([
            LinearFade('in', fade_size=(0.1, 0.2)),
            LinearFade('out', fade_size=(0.1, 0.2))
        ], n=(1, 2))
    ], p=0.5),
    TimeStretch(rate=(0.8, 1.2)),
    PitchShift(n_steps=(-0.25, 0.25)),
    MedianFilter(window_size=(5, 10), p=0.5)
], random_state=seed)

# Apply the pipeline steps to the input WAV file and write it to the output file.
transform.apply_to_wav('in.wav', 'out.wav')

generate_from_wav(self, source, n=1)¶

Runs the quantifier steps on a provided input WAV file and returns a numpy.ndarray.

Parameters

source (str, Path or path-like) – Path to the input WAV file.
n (int $> 0$) – Number of augmented versions of the source signal to generate.

Returns

The augmented versions (or version if n=1) of the source signal.

Return type

List[numpy.ndarray] or numpy.ndarray

Example:

from sigment import *

# Create a pipeline of multiple OneOf quantifiers.
transform = Pipeline([
    OneOf([
       EdgeCrop(side='start', crop_size=(0.04, 0.08)),
       EdgeCrop(side='end', crop_size=(0.04, 0.08))
    ]),
    OneOf([
        LinearFade(direction='in', fade_size=(0.02, 0.05)),
        LinearFade(direction='out', fade_size=(0.02, 0.05))
    ])
])

# Generate 5 augmented versions of the signal data from 'signal.wav' as numpy.ndarrays.
Xs_transform = transform.generate_from_wav('signal.wav', n=5)

Transformations¶

Quantifiers¶

Quantifier (Base)

Pipeline

Sometimes

Some

One