TorchDSP / torchsig

TorchSig is an open-source signal processing machine learning toolkit based on the PyTorch data handling pipeline.
MIT License
155 stars 37 forks source link

Rework of frequency shift avoid aliasing transform #210

Closed MattCarrickPL closed 11 months ago

MattCarrickPL commented 12 months ago

Frequency shift avoid aliasing has been reworked. Previously included two additional LPF's, one after interpolation and one before decimation. The two additional LPF's have been removed because the necessary filtering is already implemented in the sample rate change.

The decimation has been reworked. Previously used scipy's resample_poly() and does a decimate by 2. The problem was that it still had a large amount of energy aliasing from +fs/2 to -fs/2. The new filter uses a lower cutoff frequency which allows better rejection of aliased energy at the cost of distorting the passband of the signal.

Implements #199 (link includes design notes and commentary on changes)