Hey Remothey.

You can do this, but I'm wondering if you have considered binning the samples into discrete bins and then calculating the standard deviation of each bin? The interpolation idea is a good one (especially if the behavior is highly non-linear).

Basically the question you need to answer is what exactly you want to use the measure for in terms of the analysis and also in terms of the inference (and the associated question you are trying to answer).

If the value is good enough to answer the question then so be it, but if the value doesn't have a good enough representation for answering the question, then you will need another measure, or more data.