Numeric Pattern Matching

Question

The project I am studying requires some numerical pattern matching. My searches did not display many relevant queries, since most of the results are usually related to matching text patterns. The idea is that we will have certain wave patterns that we will need to view and try to match the input data and the wave database that we will build. Here is an example of one of the wave patterns that we need to match.

alt text http://tmp.stayhealthy.com/wave.png

There is clearly a pattern there, but the peaks will not have the same values, but the general shape of the wave iterations will be very similar. Does anyone have any tips on how to store and then match these patterns, and / or other search terms that I can use to find more information on the topic of pattern matching?

Thanks Tim.

Answer 1

FFT is useful for general data sequence analysis.

If you need to match sequences, then phase correlation is much more efficient than linear correlation. The wikipedia example is for 2D image analysis, but it can also be used in 1D.

Answer 2

Check the (discrete) Fourier transform, which can extract the time series frequency.

(And Excel has a built-in fast Fourier transform function.)

Answer 3

You should start reading about correlation .

The easiest way to start is with Linear Correlation , but get good statistics. Visual statistics seem like a very good start .

---

I know there is a Visual Studio statistical analysis package, but I can’t find more information right now.

---

Depending on your exact requirements, you can look in the R programming language , which is specially designed for statistical analysis and / or the SciPy Python library

It seems you have a simple simple variable linear correlation, if instead of showing the timeline, you do a scatter plot with X, when Series1 and Y are Series2, you will see a packed cloud of dots with a downward slope. This is a linear correlation.

Answer 4

Have you watched the Wavelets ?

Answer 5

I think it will be much easier for you to define such patterns in the temporal representation of your signal, rather than trying to directly use Fourier analysis. I made my Ph.D. on a similar problem in the context of diffraction data. The software that I used is available here .