Summary: | This paper addresses the weakness of ensemble empirical mode
decomposition approach in extracting the components of a time series
signal data. In general, this approach provides non-independent
component. The existing approach using cluster analysis provided an improvement yet not perfect. We then do a modification to reach
components with two main characteristics. First, the components
should reflect the true patterns. Second, the components are
independent among the others as much as possible. By a small
empirical study, we observe the modification we propose produces
better results than the existing approaches.
|