Study of boundary conditions in the iterative filtering method for the decomposition of nonstationary signals (Q1989171): Difference between revisions

The paper addresses the problem of choice of boundary condition in iterative filtering based technique of computing the intrinsic mode functions (IMFs). The authors show that each of the choices considered gives rise to certain type structured matrix whose spectral properties are determined in order to establish the claims of convergence. The work is technically sound and provides sufficient examples to demonstrate the ideas presented. In particular, the authors conclusively answer the first two of three questions they pose in the beginning: -- Does IF converge also for other kinds of BCs? -- Given a signal extended artificially outside the boundaries in a certain way, how do the errors introduced outside the boundaries effect the decomposition inside the signal in the iterations? -- Given a compactly supported signal what is the best choice in terms of BCs? Given that virtually nothing is known about \(\ddot{\mathbf{s}}_{\text{out}}\), it is indeed remarkable that the heuristic approach taken to estimate \(\mathbf{s}^{\mathcal{BC}}_{\text{out}}-\ddot{\mathbf{s}}_{\text{out}}\) did not get violated in the examples presented (of which one of them was real world data). However, in the examples there is really no unifying idea as to why one should expect a particular type of BC to give better results for a certain case. Can this not be held against the usefulness of the idea of IMFs? I think the authors should reflect a bit more on this issue. In general, the manuscript is well-written.