mlsacheck

class diffsptk.MLSADigitalFilterStabilityCheck(cep_order: int, *, alpha: float = 0, pade_order: int = 4, strict: bool = True, threshold: float | None = None, fast: bool = True, n_fft: int = 256, warn_type: str = 'warn', mod_type: str = 'scale', device: device | None = None, dtype: dtype | None = None)

See this page for details.

Parameters:

cep_orderint >= 0

The order of the mel-cepstrum, \(M\).

alphafloat in (-1, 1)

The frequency warping factor, \(\alpha\).

pade_orderint in [4, 7]

The order of the Pade approximation.

strictbool

If True, prioritizes maintaining the maximum log approximation error over MLSA filter stability.

thresholdfloat > 0 or None

The threshold value. If None, it is automatically computed.

fastbool

Enables fast mode (do not use FFT).

n_fftint > M

The number of FFT bins. Used only in non-fast mode.

warn_type[‘ignore’, ‘warn’, ‘exit’]

The warning type.

mod_type[‘clip’, ‘scale’]

The modification method.

devicetorch.device or None

The device of this module.

dtypetorch.dtype or None

The data type of this module.

References

[1]

S. Imai et al., “Mel log spectrum approximation (MLSA) filter for speech synthesis,” Electronics and Communications in Japan, vol. 66, no. 2, pp. 11-18, 1983.

forward(mc: Tensor) → Tensor

Check the stability of the MLSA digital filter.

Parameters:

mcTensor [shape=(…, M+1)]

The input mel-cepstrum.

Returns:

outTensor [shape=(…, M+1)]

The modified mel-cepstrum.

Examples

>>> import diffsptk
>>> import torch
>>> mlsacheck = diffsptk.MLSADigitalFilterStabilityCheck(
...     cep_order=4, alpha=0.1, warn_type="ignore"
... )
>>> c1 = torch.tensor([1.8963, 7.6629, 4.4804, 8.0669, -1.2768])
>>> c2 = mlsacheck(c1)
>>> c2
tensor([ 1.3336,  1.7537,  1.0254,  1.8462, -0.2922])

diffsptk.functional.mlsacheck(c: Tensor, *, alpha: float = 0, pade_order: int = 4, strict: bool = True, threshold: float | None = None, fast: bool = True, n_fft: int = 256, warn_type: str = 'warn', mod_type: str = 'scale') → Tensor

Check the stability of the MLSA digital filter.

Parameters:

cTensor [shape=(…, M+1)]

The input Mel-cepstrum.

alphafloat in (-1, 1)

The frequency warping factor, \(\alpha\).

pade_orderint in [4, 7]

The order of the Pade approximation.

strictbool

If True, prioritizes maintaining the maximum log approximation error over MLSA filter stability.

thresholdfloat > 0 or None

The threshold value. If None, it is automatically computed.

fastbool

Enables fast mode (do not use FFT).

n_fftint > M

The number of FFT bins. Used only in non-fast mode.

warn_type[‘ignore’, ‘warn’, ‘exit’]

The warning type.

mod_type[‘clip’, ‘scale’]

The modification method.

Returns:

outTensor [shape=(…, M+1)]

The modified mel-cepstrum.

See also

mglsadf imglsadf