Compute
$$Y_t = \sum_{k=-q}^p A_k X_{t-k}$$
where \(X_t\) is a stationary multivariate time series and \((A_k)_{-q \leq k \leq p}\) is a filter.
Note that this procedure can also be used for deconvolution
$$\tilde X_t = \sum_{k=-q}^p X_{t+k}$$ A_k'
but therefore in place of A one should it's transposition and inversed time, i.e.
t(rev(A)).
filter.process(X, A)process
time-domain operator series
Multivariate linear process