espnet2.train.preprocessor.framing

source

espnet2.train.preprocessor.framing(x, frame_length: int = 512, frame_shift: int = 256, centered: bool = True, padded: bool = True)

Segment an input array into overlapping frames.

This function divides the input array x into overlapping frames of specified length and shift. It can also pad the input and center the frames based on the provided parameters.

• Parameters:
  • x (np.ndarray) – The input array to be framed.
  • frame_length (int , optional) – The length of each frame. Must be a positive integer. Defaults to 512.
  • frame_shift (int , optional) – The number of samples to shift for the next frame. Must be greater than 0. Defaults to 256.
  • centered (bool , optional) – If True, pads the input array on both sides to center the frames. Defaults to True.
  • padded (bool , optional) – If True, pads the input array to ensure an integer number of frames can be created. Defaults to True.
• Returns: A 2D array where each row represents a frame of frame_length samples.
• Return type: np.ndarray
• Raises:
  • ValueError – If the input array size is zero, if
  • frame_length –
  • than the input length**, or** if frame_shift is less than or –
  • equal to 0. –

Examples

>>> x = np.random.rand(1024)  # Sample input
>>> frames = framing(x, frame_length=512, frame_shift=256)
>>> frames.shape
(3, 512)  # Three frames of length 512

NOTE

The output array will have shape (num_frames, frame_length), where num_frames is calculated based on the input size, frame length, and frame shift.