Need to understand Fast Fourier Transformation output

ssudeesh2000ssudeesh2000 Member Posts: 2 Contributor I
edited November 2018 in Help
Hi,

I have tried using Fast Fourier Transformation. I am getting dim1 and dim2 as two columns in output. How do I infer the output? What does this mean? I have read in a paper that they have created 32 FFT coefficients using FFT transformation. How can I build that? Thanks in advance. Sudeesh

Answers

  • prendosoprendoso Member Posts: 1 Contributor I
    Hi,
    What are you trying to do. I've not used the fft stuff in rapid miner. However, looking at the description it appears that its merely converting an input time series into the frequency components i.e. if you have a set of data in the time domain, then this set of data can be transformed to provide the constituent frequencies. This can be done analytically, or discretely - the discrete Fourier transform. The fast Fourier transform is merely a neat way to speed up the process that was discovered by a couple of guys in the 60's (I think, Cooley and Tukey).

    Forgive me if you know all this stuff.

    NB: the FFT will contain a real and imaginary component.
  • renewrenew Member Posts: 3 Contributor I

    Hi, i also want to do Fourier Transform on vibration data (time series), but also have no idea, how the output is to be interpretet. I also get dim_1 (real and imaginary value type) and dim_2 (only real value). There is no labeling on the axis and the extension documentation does not give any hint on this.

     

    Did anyone perform a succesfull fourier transform and can help us?

  • sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager

    hello @renew - the Time Series Extension is undergoing a complete rebuild under the excellent leadership of @tftemme and is currently in alpha testing.  Please see recent blog post here: http://community.rapidminer.com/t5/Community-Blog/Time-Series-Extension-Release-of-the-Alpha-Version-0-1-2/ba-p/41717  Perhaps you should give this a try?

     

    Scott

  • tftemmetftemme Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, RMResearcher, Member Posts: 164 RM Research

    Hey @renew and @ssudeesh2000. As Scott is correctly pointing out we are currently developing the rebuild of the Time Series Extension. Unfortunately a Fast Fourier Transformation is not yet in the new Extension. But it is definetly a must have and on the list for the next versions.

     

    For the old Series Extension, the _dim_1 Attribute is the Amplitude of the Fourier coefficients and probably the Attribute you want to have. Unfortunately I am not sure why you have also imaginary value type (although @prendoso is correct that a correct fourier transformation has imaginary coefficients), cause I can't reproduce it with my current setup here. 
    The _dim_2 Attribute is the phase of the corresponding coefficient. (The values should vary in a range of pi (so for example for me between 0 and 3.141)).

Sign In or Register to comment.