stft#

diffsptk.STFT#: alias of ShortTimeFourierTransform

class diffsptk.ShortTimeFourierTransform(frame_length: int, frame_period: int, fft_length: int, *, center: bool = True, zmean: bool = False, mode: str = 'constant', window: str = 'blackman', norm: str = 'power', eps: float = 1e-09, relative_floor: float | None = None, out_format: str = 'power', learnable: bool | list[str] = False)[source]#

This module is a simple cascade of framing, windowing, and spectrum calculation.

Parameters:

frame_lengthint >= 1: The frame length in samples, \(L\).
frame_periodint >= 1: The frame period in samples, \(P\).
fft_lengthint >= L: The number of FFT bins, \(N\).
centerbool: If True, pad the input on both sides so that the frame is centered.
zmeanbool: If True, perform mean subtraction on each frame.
mode[‘constant’, ‘reflect’, ‘replicate’, ‘circular’]: The padding method.
window[‘blackman’, ‘hamming’, ‘hanning’, ‘bartlett’, ‘trapezoidal’, ‘rectangular’, ‘nuttall’]: The window type.
norm[‘none’, ‘power’, ‘magnitude’]: The normalization type of the window.
epsfloat >= 0: A small value added to the power spectrum.
relative_floorfloat < 0 or None: The relative floor of the power spectrum in dB.
out_format[‘db’, ‘log-magnitude’, ‘magnitude’, ‘power’, ‘complex’]: The output format.
learnablebool or list[str]: Indicates whether the parameters are learnable. If a boolean, it specifies whether all parameters are learnable. If a list, it contains the keys of the learnable parameters, which can only be “basis” and “window”.

forward(x: Tensor) → Tensor[source]#

Compute short-time Fourier transform.

Parameters:

xTensor [shape=(…, T)]: The input waveform.

Returns:

outTensor [shape=(…, T/P, N/2+1)]: The output spectrogram.

Examples

>>> x = diffsptk.ramp(1, 3)
>>> x
tensor([1., 2., 3.])
>>> stft = diffsptk.STFT(frame_length=3, frame_period=1, fft_length=8)
>>> y = stft(x)
>>> y
tensor([[1.0000, 1.0000, 1.0000, 1.0000, 1.0000],
        [4.0000, 4.0000, 4.0000, 4.0000, 4.0000],
        [9.0000, 9.0000, 9.0000, 9.0000, 9.0000]])

diffsptk.functional.stft(x: Tensor, *, frame_length: int = 400, frame_period: int = 80, fft_length: int = 512, center: bool = True, zmean: bool = False, mode: str = 'constant', window: str = 'blackman', norm: str = 'power', eps: float = 1e-09, relative_floor: float | None = None, out_format: str = 'power') → Tensor[source]#

Compute short-time Fourier transform.

Parameters:

xTensor [shape=(…, T)]: The input waveform.
frame_lengthint >= 1: The frame length in samples, \(L\).
frame_periodint >= 1: The frame period in samples, \(P\).
fft_lengthint >= L: The number of FFT bins, \(N\).
centerbool: If True, pad the input on both sides so that the frame is centered.
zmeanbool: If True, perform mean subtraction on each frame.
mode[‘constant’, ‘reflect’, ‘replicate’, ‘circular’]: The padding method.
window[‘blackman’, ‘hamming’, ‘hanning’, ‘bartlett’, ‘trapezoidal’, ‘rectangular’, ‘nuttall’]: The window type.
norm[‘none’, ‘power’, ‘magnitude’]: The normalization type of the window.
epsfloat >= 0: A small value added to the power spectrum.
relative_floorfloat < 0 or None: The relative floor of the power spectrum in dB.
out_format[‘db’, ‘log-magnitude’, ‘magnitude’, ‘power’, ‘complex’]: The output format.

Returns:

outTensor [shape=(…, T/P, N/2+1)]: The output spectrogram.