tensorplay.signal

Window functions for spectral analysis and filter design. A window is a length-M sequence of weights that tapers a finite segment of a signal, reducing the spectral leakage that comes from cutting the segment off sharply.

All window functions share a common shape of keyword arguments:

• M — the window length, a positive integer.

• sym — whether the window is symmetric (True, the default) or periodic (False). Symmetric windows are used in filter design; periodic windows are used for spectral analysis, where the window is applied to one period of a sampled signal and the first and last samples are not duplicates.

• dtype, layout, device, requires_grad — the usual tensor construction arguments, so window tensors are created directly on the desired device with the desired data type.

import tensorplay as tp
from tensorplay.signal import windows

print(windows.hann(8))
print(windows.blackman(8, sym=False))
# window tensors are ordinary tensors, so they can be moved or differentiated
w = windows.kaiser(64, beta=8.0)