mfcc#

diffsptk.MFCC#: alias of MelFrequencyCepstralCoefficientsAnalysis

class diffsptk.MelFrequencyCepstralCoefficientsAnalysis(mfcc_order, n_channel, fft_length, sample_rate, lifter=1, out_format='y', **fbank_kwargs)[source]#

See this page for details.

Parameters:

mfcc_orderint >= 1: Order of MFCC, \(M\).
n_channelint >= 1: Number of mel-filter banks, \(C\).
fft_lengthint >= 2: Number of FFT bins, \(L\).
sample_rateint >= 1: Sample rate in Hz.
lifterint >= 1: Liftering coefficient.
f_minfloat >= 0: Minimum frequency in Hz.
f_maxfloat <= sample_rate // 2: Maximum frequency in Hz.
floorfloat > 0: Minimum mel-filter bank output in linear scale.
out_format[‘y’, ‘yE’, ‘yc’, ‘ycE’]: y is MFCC, c is C0, and E is energy.

References

[1]

Young et al., “The HTK Book,” Cambridge University Press, 2006.

forward(x)[source]#

Compute MFCC.

Parameters:

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

Returns:

yTensor [shape=(…, M)]: MFCC without C0.
ETensor [shape=(…, 1)] (optional): Energy.
cTensor [shape=(…, 1)] (optional): C0.

Examples

>>> x = diffsptk.ramp(19)
>>> stft = diffsptk.STFT(frame_length=10, frame_period=10, fft_length=32)
>>> mfcc = diffsptk.MFCC(4, 8, 32, 8000)
>>> y = mfcc(stft(x))
>>> y
tensor([[-7.7745e-03, -1.4447e-02,  1.6157e-02,  1.1069e-03],
        [ 2.8049e+00, -1.6257e+00, -2.3566e-02,  1.2804e-01]])