espnet2.gan_svs.visinger2.ddsp.resample

source

espnet2.gan_svs.visinger2.ddsp.resample(x, factor: int)

Resamples the input tensor by a given factor using a convolutional approach.

This function takes a 3D tensor (batch, frame, channel) and resamples it by inserting zeros between the original samples, followed by a convolution with a Hann window to smooth the output.

• Parameters:
  • x (torch.Tensor) – A 3D tensor of shape (batch, frame, channel) representing the input signal to be resampled.
  • factor (int) – The factor by which to resample the input tensor. A factor of 2 will double the size of the output in the time dimension.
• Returns: A 3D tensor of shape (batch, factor * frame, channel) : representing the resampled signal.
• Return type: torch.Tensor

Examples

>>> x = torch.rand(4, 100, 2)  # Batch of 4 signals, each with 100 frames and 2 channels
>>> resampled = resample(x, 2)
>>> print(resampled.shape)
torch.Size([4, 200, 2])  # Output shape is now (batch, 200 frames, channel)

NOTE

The input tensor should have at least 3 dimensions. The resampling is performed by inserting zeros between samples, followed by a convolution operation with a Hann window for smoothing.