tonic.audio_transforms

Module Contents

Classes

FixLength	Fix the length of a sample along a specified axis to a given length.
Bin	Bin the given data along a specified axis at the specified new frequency.
SwapAxes	Interchange two axes of an array.
AmplitudeScale	Normalize the maximum amplitude of the input signal to a desired value, for instance the
RobustAmplitudeScale	Normalize the maximum amplitude of the input signal while eliminating the outliers, to
SOSFilter	SOS filter.
ButterFilter	Butter filter.
ButterFilterBank	Butter filter bank.
LinearButterFilterBank	Butter filter bank.
MelButterFilterBank	Butter filter bank with frequencies along the mel scale.
AddNoise	Add nose to data.

Functions

normalize(signal)

Normalize the signal.

class tonic.audio_transforms.FixLength

Fix the length of a sample along a specified axis to a given length.

Parameters:

• length (int) – Desired length of the sample

• axis (int, optional) – Dimension along which the length needs to be fixed.. Defaults to 1.

• data (np.ndarray) – data sample

Returns:

fixed length data sample

Return type:

np.ndarray

length: int

axis: int = 1

__call__(data: numpy.ndarray) → numpy.ndarray

Parameters:

data (numpy.ndarray) –

Return type:

numpy.ndarray

class tonic.audio_transforms.Bin

Bin the given data along a specified axis at the specified new frequency.

Parameters:

• orig_freq (float) – Sampling frequency of the given data stream

• new_freq (float) – Desired frequency after binning

• axis (int) – Axis along which the data needs to be binned

• data (np.ndarray) – data sample

Returns:

binned data sample

Return type:

np.ndarray

orig_freq: float

new_freq: float

axis: int

__call__(data: numpy.ndarray) → numpy.ndarray

Parameters:

data (numpy.ndarray) –

Return type:

numpy.ndarray

class tonic.audio_transforms.SwapAxes

Interchange two axes of an array.

Parameters:

• ax1 (int) – First axis

• ax2 (int) – Second axis

Returns:

array with interchanged axes

Return type:

np.ndarray

ax1: int = 0

ax2: int = 1

__call__(data: numpy.ndarray) → numpy.ndarray

Parameters:

data (numpy.ndarray) –

Return type:

numpy.ndarray

class tonic.audio_transforms.AmplitudeScale

Normalize the maximum amplitude of the input signal to a desired value, for instance the value can corespond to the mVolt equivalent of be the maximum loudness of the audio.

Parameters:

data (np.ndarray) – input single- or multi-channel signal.

Returns:

scaled version of the signal.

Return type:

np.ndarray

max_amplitude: float = 0.158

__call__(data: numpy.ndarray) → numpy.ndarray

Parameters:

data (numpy.ndarray) –

Return type:

numpy.ndarray

class tonic.audio_transforms.RobustAmplitudeScale

Normalize the maximum amplitude of the input signal while eliminating the outliers, to ensure that some very large outlier samples do not attenuate the majority of the samples.

Paremeters:

max_robust_amplitude (float): maximum amplitude of non-outlier samples of the signal after robust scaling. outlier_percent (float, optional): percentage of the outlier in the data. Defaults to 0.01.

Parameters:

data (np.ndarray) – input single- or multi-channel signal.

Returns:

scaled version of the signal.

Return type:

np.ndarray

max_robust_amplitude: float = 0.158

outlier_percent: float = 0.01

__call__(data: numpy.ndarray) → numpy.ndarray

Parameters:

data (numpy.ndarray) –

Return type:

numpy.ndarray

class tonic.audio_transforms.SOSFilter

SOS filter.

Parameters:

• coeffs – coefficients of the second order filter

• axis – Axis along with the filter needs to be applied

• https (See) –

coeffs: numpy.ndarray

axis: int

__call__(signal)

class tonic.audio_transforms.ButterFilter

Butter filter.

Parameters:

• order – Order of filter to be used

• freq – Frequency for the filter (float or (float, float))

• analog – True if analog filter

• btype – Filter type, {‘lowpass’, ‘highpass’, ‘bandpass’, ‘bandstop’}

• rectify – If true, the output is the absolute value of the filtered output

• axis – Axis along which the filter needs to be applied

• https (See) –

order: int

freq: Union[float, Tuple[float, float]]

analog: bool

btype: str

rectify: bool

axis: int

__post_init__()

__call__(data: numpy.ndarray) → numpy.ndarray

Parameters:

data (numpy.ndarray) –

Return type:

numpy.ndarray

class tonic.audio_transforms.ButterFilterBank

Butter filter bank.

Parameters:

• order – Order of filter to be used

• freq – Frequency for the filter (float or (float, float))

• rectify – If true, the output is the absolute value of the filtered output

• axis – Axis along which the filter needs to be applied

• analog – If true, the filter will be analog. False by default

• https (See) –

order: int

freq: List[Tuple[float, float]]

rectify: bool

axis: int

analog: bool = False

__post_init__()

__call__(data)

class tonic.audio_transforms.LinearButterFilterBank

Butter filter bank.

Parameters:

• order – Order of filter to be used

• low_freq – Lower/cutoff frequency the filter (float or (float, float))

• sampling_freq – Sampling frequency of the signal, also serves as higher frequency of the filter bank.

• analog – True if analog filter

• rectify – If true, the output is the absolute value of the filtered output

• axis – Axis along which the filter needs to be applied

• https (See) –

order: int = 2

low_freq: float = 100

sampling_freq: float = 16000

analog: bool = False

num_filters: int = 64

rectify: bool = True

axis: int

compute_freq_bands()

__post_init__()

__call__(data)

class tonic.audio_transforms.MelButterFilterBank

Bases: LinearButterFilterBank

Butter filter bank with frequencies along the mel scale.

Parameters:

• order – Order of filter to be used

• low_freq – Lower/cutoff frequency the filter (float or (float, float))

• sampling_freq – Sampling frequency of the signal, also serves as higher frequency of the filter bank.

• analog – True if analog filter

• rectify – If true, the output is the absolute value of the filtered output

• axis – Axis along which the filter needs to be applied

• https (See) –

static hz2mel(freq)

static mel2hz(freq)

compute_freq_bands()

class tonic.audio_transforms.AddNoise

Add nose to data.

Params:

dataset:

A dataset object that returns a tuple when iterated over the first element of which is the audio signal to be used for noise.

snr:

Desired signal to noise ratio in dB

normed:

If set to false, the signal max value will not be normalized. True by default.

dataset: Iterator

snr: float

normed: bool = True

get_noise_sample(sample_len: int) → numpy.ndarray

Get a random noise sample from the dataset.

Parameters:

sample_len (int) –

Return type:

numpy.ndarray

__call__(signal)

tonic.audio_transforms.normalize(signal)

Normalize the signal.