Applies an IIR filter to a 1D numeric signal using forward and backward passes
to eliminate phase distortions, similar to scipy.signal.filtfilt.
This function implements a pure C++ version using the Direct Form II Transposed structure,
including optional initial condition handling via steady-state initialization.
Arguments
- x
A numeric vector representing the input time series.
- b
A numeric vector of numerator (feedforward) filter coefficients.
- a
A numeric vector of denominator (feedback) filter coefficients. Must have
a[0] == 1.0.- padlen
Number of samples to extend on each edge for padding. If
-1(default), uses3 * max(length(a), length(b)).- padtype
Type of padding at the signal boundaries. One of
"constant"(default),"odd","even", or"zero".- use_zi
Logical. If
TRUE(default), use steady-state initial conditions to minimize transients.
Details
The function applies the IIR filter in the forward direction, reverses the result,
and applies the filter again, then reverses the final output. This approach
removes phase delay. Padding is used to minimize edge artifacts, and the filter
state is optionally initialized with values derived from the steady-state
step response of the filter (lfilter_zi_arma()).
References
Scipy Signal Documentation: https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.filtfilt.html
Gustafsson, F. (1996). Determining the initial states in forward-backward filtering. IEEE Transactions on Signal Processing.