How to define an envelope tube?

Hi everybody, I would appreciate ideas of how one could possibly define some sort of envelope tube. I already tried something on my own, but maybe there is someone having a brilliant idea about it? Nailing down the problem:

Given a two-dimensional signal dependent on time, say, x- and y-component of time-varying magnetic field (Bx, By), which could be visualized via plot3, a tube shall be derived from this data, enclosing it tightly. In case of only one dimension, one would possible derive this envelope (now: band-like area envelope) by interpolating over maxima and minima of the data. The tube though, would not only be pretty to visualize data, its mean is the interesting thing to use for postprocessing (keyword: Empirical Mode Decomposition, for those interested in. others: just ignore)

I've already implemented algorithms interpolating over minima and maxima of both components, meaning to represent four boundary lines of this tube. The result I didn't find as satisfying as I whished, though...

Creative task: Anyone with an elegant idea how derive such a tube from given data, perhaps even iterative methods? I would like to receive some sort of brainstorming for crazy ideas worth testing:)

Thanks to everybody participating, and, sorry for any inconsistencies in language, for German is my mother tongue;)

Answer 1

[Matt J]

You might have to better define the properties you desire in this envelope. You could consider using alpha shape generators, e.g.,

http://www.mathworks.com/matlabcentral/fileexchange/28851-alpha-shapes