How to filter a signal using transfer function in Pytorch? #25180


I am trying to implement a 1D convolution operation with F.conv1d. The current usage of this function is to provide the weights of the filter directly in the time domain. However, for some DSP purposes, it is more effective to apply the filtering process in terms of the rational transfer function of the filter. In other words, to provide the coefficients of numerator and denominator of the transfer function and the function applies the filtering process accordingly.

This is already provided in scipy as well as Matlab.

I think it is possible to do the same with Pytorch, but I am still struggling to grasp all the details to achieve this … could you please give any tips?

Many thanks in advance