|
|
Our Conclusion
after completing this project is that adaptive filters are excellent analysis tools, if designed and used correctly. Adaptive filters, given their nature, adapt to the variations within the input data, and this adaptation must be prescribed and predefined. As we quickly discovered when we tried to improve upon the non-adaptive filters that we started with, programming adaptation is a tremendously arduous task, and requires quite an excessive amount of retooling and rethinking in order to maximize the amount of information extracted from the data. Difficulty in programming adaptability seems to increase exponentially the more that the input data separates itself from the simple case. However, if crucial information cannot be gleaned from nonadaptive filtering, then an adaptive filter becomes not only highly useful, but necessary. Because of our low-level confrontation with the guts of adaptive filters, we would recommend designing one only in this case, where the need crucial information makes an adaptive filter necessary. Sequential nonadaptive filtering can provide a highly tailored analysis without nearly as much effort as the design of an adaptive filter would require. Therefore, while we praise the ability of adaptive filters such as EMD to effectively extract necessary information from data, we highly recommend first assuming the simplicity and efficiency of nonadaptive filters will suffice.
|
|
