[object Object][object Object][object Object]undefined[object Object]

• Function: Computes the Fourier transform of the input in a sliding window.

• Formula:

  • When [object Object] is set to [object Object]:

    $$X[w,m]=\sum_{k=0}^{winLength-1}window[k]*self[m*hopLength+k]*exp(-j*\frac{2{\pi}wk}{nFft})$$

  • When [object Object] is set to [object Object]:

    $$X[w,m]=\frac{1}{\sqrt{nFft}}(\sum_{k=0}^{winLength-1}window[k]*self[m*hopLength+k]*exp(-j*\frac{2{\pi}wk}{nFft}))$$

  Where:

  • FFT works on frequency [object Object].
  • [object Object] is the index of the sliding window.
  • [object Object] is a 1D or 2D tensor. When $self$ is 1D, there is only one time sequence. When [object Object] is 2D, there are multiple time sequences.
  • [object Object] is the sliding window size.
  • [object Object] is a 1D tensor, which is the window function (for example, hann_window) of STFT. Its length is [object Object].
  • [object Object] is the rotation factor.

[object Object]

Each operator has calls. First, [object Object] is called to obtain the workspace size required for computation and the executor that contains the operator computation process. Then, [object Object] is called to perform computation.

[object Object]

• Parameters

  [object Object]

• Returns:

  [object Object]: status code. For details, see .

  The first-phase API implements input parameter verification. The following errors may be thrown:

  [object Object]

[object Object]

• Parameters

  [object Object]

• Returns:

  [object Object]: status code. For details, see .

[object Object]

• The input [object Object] is different from that of the PyTorch interface. The input [object Object] of the PyTorch interface is the original input, while the input [object Object] of [object Object] is the result after the original input is padded by PyTorch.

• When the shape of the input self is [B, L], if the calculation result of the following formula is large, the calculation of the current API may time out.

  $$B * ((L - nFft) / hopLength + 1) * nFft$$

• nFft ≤ L

• winLength ≤ nFft

• normalized = True:

  $$STFT(w,m)=\frac{1}{\sqrt{N}}X[w,m]$$

• The mappings between the input and output data types of [object Object], [object Object], [object Object], and [object Object] are as follows:

  [object Object]undefined

• Deterministic computing:

  • [object Object] defaults to a deterministic implementation.

[object Object]

The following example is for reference only. For details, see .