I want to use wavelet analysis for the estimation of different time scale variability in climate time series. Is it the right approach? Is it possible to do in Matlab? What are the steps with an example?
Hi,
you might want to check out a technique called "empirical mode decomposition"; it is purely empirical but can be very powerful in decomposing a signal into deifferent additive components with different frequency ranges. You don't even need to assume stationarity!
Dear Muhammad,
we use both (Wavelet Transform and Empirical Mode Decomposition), but be advised, that they are by far NOT "no-brainers". You will need to play around with both methods to get a reasonable answer. Fourier Transform does not give you time resolution, instead use Short Term Fourier Transform or the periodogram to estimate frequency content with some time resolution (when does with frequency appear...)
WT can be used in Matlab with the wavelet toolbox (or other web-avalailable sources), EMD sources can be found with Patrick Flandrin's website: http://perso.ens-lyon.fr/patrick.flandrin/emd.html
and all the FT and STFT routines are Matlab built-ins anyways.
Have fun!
Other transforms, that you may be able to investigate for the task are: Vigner-Ville, Zolotarev. But I agree, with the former opinions: Periodicity should be treated with Fourier..however the time-to-frequency uncertainty will be unavoidable introduced.
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series