spafe.utils.cepstral

spafe.utils.cepstral.cep2spec(cep, ncep, nfreq, dct_type=2)

Reverse the cepstrum to recover a spectrum.

Parameters:

• cep (array) – cepstral data to convert to spectral data.
• ncep (int) – number of cepstrals.
• nfreq (int) – number of points to reconstruct in spectrum.
• dct_type (int) – the discrete cosine transform type. Default is 2.

Returns

2d-array spec, spectrum matrix 2d-array idctm, the IDCT matrix that spec was multiplied by to give cep.

spafe.utils.cepstral.cmn(x)

Mean normalization.

Parameters:x (array) – input data. Returns:array with the mean normalized data.

spafe.utils.cepstral.cms(x)

Mean Substraction: Centering

Parameters:x (array) – input data. Returns:array with the centered data.

spafe.utils.cepstral.cmvn(x)

Mean Variance Normalisation

Parameters:x (array) – input data. Returns:array with the mean and variance normalized data.

spafe.utils.cepstral.cvn(x)

Variance Normalisation: Standardization

Parameters:x (array) – input data. Returns:array with the variance normalized data.

spafe.utils.cepstral.deltas(x, w=9)

Calculate the deltas (derivatives) of an input sequence with a W-points window (W odd, default 9) using a simple linear slope. This mirrors the delta calculation performed in feacalc etc. Each row of X is filtered separately.

Parameters:

• x (array) – input sequence
• w (int) – window size to use in the derivatives calculation. Default is 9.

Returns:

2d-arrays containing the derivatives values.

spafe.utils.cepstral.lifter_ceps(cepstra, L=22)

Apply a cepstral lifter the the matrix of cepstra. This has the effect of increasing the magnitude of the high frequency DCT coeffs.

Parameters:

• cepstra (np.array) – the matrix of mel-cepstra, will be numframes * numcep in size.
• L (int) – the liftering coefficient to use. Default is 22. L <= 0 disables lifter.

Returns:

liftered cepstra.

spafe.utils.cepstral.spec2cep(spec, ncep=9, dct_type=2)

Calculate cepstra from spectral samples (in columns of spec).

Parameters:

• spec (array) – spectral data to convert to cepstral.
• ncep (int) – number of cepstrals. Default is 9.
• dct_type (int) – the discrete cosine transform type. Default is 2.

Returns

2d-array ncep, cepstral rows 2d-array dctm, the DCT matrix that spec was multiplied by to give cep.