I am trying to implement scipy.signal.filtfilt function in c++ and I am wondering if there is already an implementation available of this?
3 Answers
filtfilt applies an IIR filter twice, once going forward, once going backward. The nontrivial part is how to initialize the IIR filter at the boundaries.
As a starting point, look at how scipy.signal.filtfilt does it. Here is the code: https://github.com/scipy/scipy/blob/master/scipy/signal/signaltools.py#L3870
You might also find it useful to look at the source code for Octave's filtfilt (M code): https://sourceforge.net/p/octave/signal/ci/default/tree/inst/filtfilt.m
To reproduce filtfilt in C++, you need a C++ implementation of IIR filtering to take the role of scipy's lfilter plus some boundary handling logic. I don't know about an existing C++ implementation of filtfilt, but at least for the default method='pad' logic, the core computation seems simple enough to consider porting directly.
Scipy's filtfilt is similar to Matlab's filtfilt.
A question for MATLAB's filtfilt was previously asked
An implementation for the same was previously shared on Stackoverflow by @darien-pardinas
Do note I say similar because as mentioned by @paco-wong
The difference is in the default padding length. In matlab's filtfilt, it is 3*(max(len(a), len(b)) - 1), and in scipy's filtfilt, it is 3*max(len(a), len(b)).
So you will have to account for that
Related