yingram

class diffsptk.Yingram(frame_length: int, sample_rate: int = 22050, lag_min: int = 22, lag_max: int | None = None, n_bin: int = 20)

Pitch-related feature extraction module based on YIN.

Parameters:

frame_lengthint >= 1

The frame length in samples, \(L\).

sample_rateint >= 8000

The sample rate in Hz.

lag_minint >= 1

The minimum lag in points.

lag_maxint < L

The maximum lag in points.

n_binint >= 1

The number of bins to represent a semitone range.

References

[1]

A. Cheveigne and H. Kawahara, “YIN, a fundamental frequency estimator for speech and music,” The Journal of the Acoustical Society of America, vol. 111, 2002.

[2]

H. Choi et al., “Neural analysis and synthesis: Reconstructing speech from self-supervised representations,” arXiv:2110.14513, 2021.

forward(x: Tensor) → Tensor

Compute the YIN derivatives from the waveform.

Parameters:

xTensor [shape=(…, L)]

The framed waveform.

Returns:

outTensor [shape=(…, M)]

The Yingram.

Examples

>>> x = diffsptk.nrand(22050)
>>> frame = diffsptk.Frame(2048, 441)
>>> yingram = diffsptk.Yingram(2048)
>>> y = yingram(frame(x))
>>> y.shape
torch.Size([51, 1580])