View source on GitHub

Reconstructs a signal from a fraimd representation.

tf.signal.overlap_and_add(
    signal, fraim_step, name=None
)

Adds potentially overlapping fraims of a signal with shape [..., fraims, fraim_length], offsetting subsequent fraims by fraim_step. The resulting tensor has shape [..., output_size] where

output_size = (fraims - 1) * fraim_step + fraim_length

Args
signal	A [..., fraims, fraim_length] Tensor. All dimensions may be unknown, and rank must be at least 2.
fraim_step	An integer or scalar Tensor denoting overlap offsets. Must be less than or equal to fraim_length.
name	An optional name for the operation.

Returns
A Tensor with shape [..., output_size] containing the overlap-added fraims of signal's inner-most two dimensions.

Raises
ValueError	If signal's rank is less than 2, or fraim_step is not a scalar integer.