amplitude_warp

Name	Description
`AmplitudeWarp`	Amplitude warping layer

Name

Description

AmplitudeWarp

Amplitude warping layer

AmplitudeWarp(
    sample_rate: float = 1,
    frequency: float | tuple[float, float] = 100,
    amplitude: float | tuple[float, float] = 0.1,
    **kwargs={},
)

Apply amplitude warping to the 1D input. Time points are first generated at given frequency resolution with amplitude picked from uniform distribution. These points are then interpolated to match the input duration and multiplied to the input.

Example:

sample_rate = 100 # Hz
duration = 3*sample_rate # 3 seconds
sig_freq = 10 # Hz
sig_amp = 1 # Signal amplitude
noise_freq = (1, 2) # Noise frequency range
amplitude = (0.5, 2) # Noise amplitude range
x = sig_amp*np.sin(2*np.pi*sig_freq*np.arange(duration)/sample_rate).reshape(-1, 1).astype(np.float32)
x = keras.ops.convert_to_tensor(x)
import helia_edge as helia

lyr = helia.layers.preprocessing.AmplitudeWarp(
    sample_rate=sample_rate,
    frequency=noise_freq,
    amplitude=amplitude,
)
y = lyr(x)
plt.plot(x.numpy())
plt.plot(y.numpy())
plt.show()

Base class: BaseAugmentation1D.

Inherited from BaseAugmentation: augment_masks(), augment_sample(), augment_targets(), batch_augment(), call().

Parameters

Parameters of AmplitudeWarp
Name	Type	Default	Description
`sample_rate`	`float`	1	Sample rate of the input.
`frequency`	`float \| tuple[float, float]`	100	Frequency of the warping in Hz. If tuple, frequency is randomly picked between the values.
`amplitude`	`float \| tuple[float, float]`	0.1	Amplitude of the warping. If tuple, amplitude is randomly picked between the values.

noise_type: str

sample_rate: float = sample_rate

frequency: tuple[float, float] = parse_factor(frequency, min_value=None, max_value=sample_rate / 2, param_name='frequency')

amplitude: tuple[float, float] = parse_factor(amplitude, min_value=0, max_value=None, param_name='amplitude')

get_random_transformations(input_shape: tuple[int, int, int]) -> dict

Generate noise distortion tensor

Parameters

Parameters of get_random_transformations
Name	Type	Default	Description
`input_shape`	`tuple[int, ...]`	Required	Input shape.

Returns

Returns of get_random_transformations
Value	Type	Description
`dict`	`dict`	Dictionary containing the noise tensor.

augment_samples(inputs) -> keras.KerasTensor

Augment all samples in the batch as it’s faster.

get_config()

Serialize the layer configuration to a JSON-compatible dictionary.

amplitude_warp

Amplitude Warp Layer