fftcep#

class diffsptk.CepstralAnalysis(cep_order, fft_length, n_iter=0, accel=0)[source]#

See this page for details.

Parameters:

cep_orderint >= 0: Order of cepstrum, \(M\).
fft_lengthint >= 2M: Number of FFT bins, \(L\).
n_iterint >= 0: Number of iterations.
accelfloat >= 0: Acceleration factor.

References

[1]

S. Imai et al., “Spectral envelope extraction by improved cepstral method,” IEICE trans, vol. J62-A, no. 4, pp. 217-223, 1979 (in Japanese).

forward(x)[source]#

Estimate cepstrum from spectrum.

Parameters:

xTensor [shape=(…, L/2+1)]: Power spectrum.

Returns:

outTensor [shape=(…, M+1)]: Cepstrum.

Examples

>>> x = diffsptk.ramp(19)
>>> stft = diffsptk.STFT(frame_length=10, frame_period=10, fft_length=16)
>>> fftcep = diffsptk.CepstralAnalysis(3, 16)
>>> c = fftcep(stft(x))
>>> c
tensor([[-0.9663,  0.8190, -0.0932, -0.0152],
        [-0.8539,  4.6173, -0.5496, -0.3207]])

diffsptk.functional.fftcep(x, cep_order, n_iter=0, accel=0)[source]#

Estimate cepstrum from spectrum.

Parameters:

xTensor [shape=(…, L/2+1)]: Power spectrum.
cep_orderint >= 0: Order of cepstrum, \(M\).
n_iterint >= 0: Number of iterations.
accelfloat >= 0: Acceleration factor.

Returns:

outTensor [shape=(…, M+1)]: Cepstrum.

fftcep#

This Page