pnorm#

class diffsptk.MelCepstrumPowerNormalization(cep_order: int, alpha: float = 0, ir_length: int = 128)[source]#

See this page for details.

Parameters:

cep_orderint >= 0: The order of the cepstrum, \(M\).
alphafloat in (-1, 1): The frequency warping factor, \(\alpha\).
ir_lengthint >= 1: The length of the impulse response.

forward(x: Tensor) → Tensor[source]#

Perform mel-cepstrum power normalization.

Parameters:

xTensor [shape=(…, M+1)]: The input mel-cepstrum.

Returns:

outTensor [shape=(…, M+2)]: The log power and power-normalized mel-cepstrum.

Examples

>>> x = diffsptk.ramp(1, 4)
>>> pnorm = diffsptk.MelCepstrumPowerNormalization(3, alpha=0.1)
>>> y = pnorm(x)
>>> y
tensor([ 8.2942, -7.2942,  2.0000,  3.0000,  4.0000])

diffsptk.functional.pnorm(x: Tensor, alpha: float = 0, ir_length: int = 128) → Tensor[source]#

Perform cepstrum power normalization.

Parameters:

xTensor [shape=(…, M+1)]: The input mel-cepstrum.
alphafloat in (-1, 1): The frequency warping factor, \(\alpha\).
ir_lengthint >= 1: The length of the impulse response.

Returns:

outTensor [shape=(…, M+2)]: The power-normalized cepstrum.